Eindhoven University of Technology MASTER Development and testing of a scaled tractor for use in the TruckLab Smeulders, G.A.M.

(1)

Eindhoven University of Technology

MASTER

Development and testing of a scaled tractor for use in the TruckLab

Smeulders, G.A.M.

Award date:

2020

Link to publication

Disclaimer

This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

(2)

Development and testing of a scaled tractor for use in the TruckLab

Master Thesis Report

DC 2020.050

G.A.M. Smeulders (0842719) Supervisors:

Dr. Ir. I.J.M. Besselink (TU/e) Prof. dr. Henk Nijmeijer (TU/e) Department: Mechanical Engineering Masters Program: Systems and Control Research group: Dynamics and Control

Eindhoven, Saturday 8

^th

August, 2020

(3)

(4)

Declaration concerning the TU/e Code of Scientific Conduct for the Master’s thesis

I have read the TU/e Code of Scientific Conductⁱ.

I hereby declare that my Master’s thesis has been carried out in accordance with the rules of the TU/e Code of Scientific Conduct

Date

………..…………..

Name

………..…………..

ID-number

………..…………..

Signature

………..…………..

Submit the signed declaration to the student administration of your department.

i See: http://www.tue.nl/en/university/about-the-university/integrity/scientific-integrity/

The Netherlands Code of Conduct for Academic Practice of the VSNUcan be found here also.

More information about scientific integrity is published on the websites of TU/e and VSNU

08/08/2020

G.A.M. Smeulders

0842719

(5)

(6)

Abstract

The TruckLab located in the Automotive Engineering Science Laboratory at the TU/e is used for experiments with scaled tractor semitrailer combinations. The focus is on autonomous manoeu- vring at distribution centers.

The existing version of the scaled tractor has several problems such as low resolution and quality of actuators resulting in low steering and propulsion accuracy and repeatability. Free play and non-durable materials in the steering and propulsion system result in low precision and wear over time. The limited space available also gives problems to fit components on the tractor.

Therefore this project focuses on the development and testing of an improved scaled tractor. The problems have been categorized for the main subsystems of the tractor. The most important subsystems are the steering system, the propulsion system and the computing/electronic system including sensors. These subsystems are connected by a chassis. To be able to improve the scaled tractor, other remote controlled scaled vehicles and robots as well as components have been researched.

Based on the outcome of this study and the system requirements for the new tractor, components have been selected and manufactured to create an improved tractor. Selected components include Dynamixel servo’s for actuation, a Raspberry Pi with camera, lidar and modular plates which offer adaptability. These components are normally part of a unicycle robot called Turtlebot3 Waffle Pi. Components from Turtlebot3 have been combined with components for remote controlled toy tractors and self made components. Robot Operating System (ROS) software is used for extra software tools and interacts with MATLAB/Simulink to control the new tractor.

The component integration in the newly developed metal chassis is discussed. Several options for a new steering mechanism have been evaluated. Drawings have been used as a method to evaluate integration options upfront.

The control of the tractor is developed. The main aspects are control of the steering system and control of the propulsion system. The propulsion system uses individual motors for each rear wheel, that allows for individual control of the rear wheel velocities. The steering system has been calibrated by measuring the steer angles of the front wheels and comparing those against the values predicted by the Ackermann steering geometry. Ackermann steering geometry is desired as it results in low amount of lateral slip of the tires which makes the vehicle more predictable during cornering. A polynomial relation has been determined between the steer servo position and the steer angle. For control of the tractor several software packages interact. Some adaptations needed to be made to the Turtlebot3 firmware, these adaptations are required due to the addition of an extra Dynamixel servo for actuating the steering system.

The new version of the tractor is tested by doing several experiments. These include the compar- ison of the driven path of the scaled tractor, to the path predicted by a kinematic model.

On the basis of the experimental results and the properties of the new tractor, conclusions are drawn. The steering system and the propulsion system of the new tractor have improved compared to the systems in the previous tractor. Most of the system requirements have been met. Because the new tractor is the result of a complete revision, also other problems have been successfully addressed. The new tractor has a simple system architecture as a result of only using essential components. Recommendations are given for further improvement of the tractor. More research is needed regarding the correspondence of a path driven by the new tractor and a path predicted by a (kinematic) model.

(7)

(8)

Preface

This Master thesis describes the work which has been done during my graduation project for the TruckLab located in the AES lab at the TU/e. There are some people which I would like to thank for their support during this project.

First of all I would like to thank my supervisor Dr. Ir. I.J.M. Besselink for his guidance and effective explanation and inspiration throughout the project. Thanks to my mentor Prof. dr.

Henk Nijmeijer for his feedback during the progress meetings, which has challenged me to keep improving.

Thanks to Erwin Meinders for all his help and valuable ideas on multiple occasions and for creating the nice atmosphere in the AES lab. Thanks to Wietse Loor for sharing his knowledge on electronics and other support. Thanks to Pieter van Hoof for his great manufacturing skills that made the realization of the new tractor possible and for his informative explanation about the operation of the machines and the tools.

Thanks to Dan Cristian Chirascu for his software assistance and insights, while he was also working in the TruckLab for his own graduation project. I would also like to thank his supervisor Dr. Ir. Ion Barosan for his input during the TruckLab meetings.

Thanks to Dr. Ir. Rudolf Huisman and his colleagues Paul van Alphen and Roger Bosmans at DAF for their help and expertise in initializing the process of 3D printing a DAF cabin for the tractor. Also thanks to Jasper Sterk for sharing his expertise on 3D printing.

Furthermore I would like to thank all the members of ’the famous AES lab’ for their teamspirit and the nice conversations and fun we have had during our time in the AES lab.

Special thanks to my girlfriend Nikky for all the wonderful times we have had already since we have met during Introweek at TU/e. Also thanks to her family for their support.

I would finally like to thank my parents Ad Smeulders and Miriam Smeulders and my sister Bronwyn Smeulders for all their support during my studies at the TU/e.

(9)

(10)

List of Symbols

ψ˙ yaw velocity of the tractor in kinematic model

˙

xg velocity of the tractor in x-direction of the global reference frame at the rear axle in kinematic model

˙

yg velocity of the tractor in y-direction of the global reference frame at the rear axle in kinematic model

ωz yaw velocity of the tractor

ω_ref reference yaw velocity for kinematic model C center of rotation during cornering

d_f distance between the front wheels d_w distance between the rear wheels

J_vf front joint value giving the position of the steer servo Ko proportional gain acting on orientation error

Kp proportional gain acting on positional error Ky proportional gain acting on yaw velocity error L wheelbase of the scaled tractor

R radius of the curve rw radius of the rear wheels

vm longitudinal velocity of the tractor in kinematic model vcmd,l velocity of left rear wheel expressed in 0.229 rev/min vcmd,r velocity of right rear wheel expressed in 0.229 rev/min v_ref reference velocity for kinematic model

v_x,l linear velocity of left rear wheel v_x,r linear velocity of right rear wheel v_x vehicle longitudinal velocity

x_g x-position mid rear axle in kinematic model in global reference frame x_ref reference x-position for kinematic model

yg y-position mid rear axle in kinematic model in global reference frame yref reference y-position for kinematic model

δ steer angle of the tractor

δl steer angle of the left front wheel of the tractor δm steer angle of the tractor in kinematic model δr steer angle of the right front wheel of the tractor

δf b,o steer angle resulting from feedback control on orientation δf b,p steer angle resulting from feedback control on position δf b,y steer angle resulting from feedback control on yaw velocity δ_{f b} steer angle resulting from feedback control

δ_{f f} steer angle resulting from feed forward control δ_ref reference steer angle for kinematic model

ψ orientation of the tractor in the global reference frame in kinematic model ψ_ref reference orientation for kinematic model

θ_tp angle between the vector which points from mid rear axle to the reference point on the path and the vector through the center line of the vehicle

(13)

(14)

Chapter 1

Introduction

1.1 Background of the project and past results

Tractor semitrailer combinations have an important role in the logistics sector. They are often used because of the easy separation of the tractor from the semitrailer. Tractor semitrailer combinations can however be a challenge to drive. Especially at distribution centres where complex manoeuvres such as docking are needed to load and unload goods.

Smart Mobility Labs such as at KTH Stockholm [30] are used by researchers to experiment with tractor semitrailer combinations in a safe manner by using scaled vehicles in a controlled laboratory environment. In the Smart Mobility Lab at KTH it is studied how traffic congestion and emissions caused by tractor semitrailer combinations can be reduced. Fuel consumption and traffic congestion can be reduced by using platoons of Heavy Duty Vehicles according to [30]

and [31]. The short distances in between platooning vehicles requires automation of the driving task. At KTH it is further explored how the number of stops of Heavy Duty Vehicles can be reduced by using real time traffic data, again to reduce the fuel consumption and to increase traffic flow. Autonomous driving systems for reversing an articulated vehicle are also researched at KTH.

The TruckLab [2], [5] is a research setup located in the Automotive Technology laboratory at TU/e. The TruckLab is used to investigate autonomous driving of tractor semitrailer combinations. The focus is to autonomously perform manoeuvres which occur at distribution centers.

These manoeuvres include backing up the tractor semitrailer combination towards a docking bay, parallel parking and following another tractor semitrailer combination. Being able to perform these manoeuvres autonomously can give benefits in the logistics sector. To study the development of these autonomous manoeuvres in a safe and cost effective way, scaled tractor semitrailer combinations and a scaled distribution centre are used at TU/e.

One of the advantages of automated driving is that it reduces the reliance on the skills of truck drivers. Drivers have some limitations, for instance there inability to see the area behind the vehicle when driving in reverse. These limitations can lead to accidents, which may result in damage to the trucks or the distribution centre. In hazardous situations people can also get hurt. Automation of complex manoeuvres not only increases safety, it can also reduce the time needed to complete manoeuvres. This gives economical benefits as congestion at logistic centres can thereby be reduced. Further economical benefits can be obtained as automation can give the ability to dock, load and unload trailers without a driver being present in the vehicle.

The hardware of the TruckLab currently consists of two model tractor semitrailers. These can be controlled by a Simulink control system running on a PC. The two subsystems which are actuated are the steering system and the propulsion system of the tractor. An overhead camera

(15)

CHAPTER 1. INTRODUCTION

system is used to determine the position and orientation of the tractor and the semitrailer by ob- serving Aruco markers. Communication between the tractors and the PC which runs the Simulink program is enabled by a Wi-Fi communication system. The existing tractors are equipped with a Olimex micro-controller to process the received control output which is created in Simulink.

Although the current hardware has allowed to do some experiments with control algorithms, it has been observed that the hardware is not good enough to perform the autonomous manoeuvres with the desired accuracy and precision.

According to [4] the path following controller which is used for autonomous parallel parking at the TruckLab, is expected to track the reference path accurately, with an error less than 1 mm.

However the maximum distance error to the reference path is still 0.17 m in the parallel parking test done in [4].

At the end of an autonomous docking manoeuvre at the TruckLab described in [3] the observed lateral distance error to the reference path is in the order of millimeters. However due to measurement uncertainties in the overhead camera system, this observed error has an uncertainty of a couple of centimeters. Therefore it cannot be concluded if the result is good enough. The articula- tion angle is -5.6 degrees at the end of the docking manoeuvre, while ideally it should be 0 degrees.

1.2 Problem definition and research objective

To be able to carry out experiments in the TruckLab which give reliable and repeatable test results and to perform autonomous manoeuvres with the desired accuracy and precision, the hardware of the TruckLab needs to be improved. Therefore this project will focus on improving the scaled tractors which are used in the TruckLab. The improvement which is gained needs to be demonstrated by means of experiments done with the new hardware and will be compared with the current hardware.

The overhead camera system and the Wi-Fi based communication system are not physically attached to the scaled tractor semitrailer and the problems related to these two systems, will be addressed in a master thesis [29] by Dan Cristian Chirascu.

In previous projects involving the TruckLab, Simulink control systems have been implemented to fulfill autonomous manoeuvres with the scaled tractor semitrailers. The problems which need to be solved and the choices which need to be made regarding these systems have been partially addressed already. Although more research in this area is needed, the developed Simulink control systems cannot be properly tested with the current hardware of the TruckLab. The problems which have been identified related to the overhead camera system, the Wi-Fi based communication system and Simulink are listed in Appendix A.

In the next section the TruckLab is divided into subsystems. For each of these subsystems the associated problems and limitations are discussed. This list of problems and limitations has been compiled on the basis of observations described in [1], [2], [3], [4] and [5]. Based on this information the research objective for this graduation project will be formulated.

1.2.1 Problem definition

This section describes the problems which should be solved. Some other problems will not be addressed by this project, they are described in Appendix A.

(16)

Tractor system in general

• The tractor contains too many unnecessary parts, for example a fake spare wheel and a sound and light module. All these additional components make it difficult to fit and reach the components on the 1/13.3 scaled tractor which are necessary to enable the research objectives of the TruckLab.

• Certain components are fragile because they are made of plastic.

Steering system

The layout of the existing tractor is shown in Figure1.1. The figure indicates the location of the most important components of the tractor and is useful to understand where problems occur.

Figure 1.1: Layout of the old tractor

• The steering servo which actuates the steering system with a frequency of 50 Hz has a resolution of around 1.1 degrees (47 steps and 52 degrees of range). This low resolution results from the small PWM range which is used to control the steering servo.

• The maximum steering angle is limited to 26 degrees due to components of the steering system contacting each other, this results in a larger turning radius than desired. The maximum steering angle of a real tractor is approximately 49 degrees according to [2].

• The steering servo is not equipped with position feedback, therefore the steering system can only be controlled in open loop.

• A free play of around 3 degrees is present in the steering system. This results in the steering angle output not being reliable nor repeatable. This free play occurs in the connection between the steering servo and the chassis, the connection between the wheel hub and the wheel axle and the connection between the wheel axle and the steering knuckle.

(17)

• The relation between the PWM input and the steering angle output differs between the two scaled trucks.

• The steering system suffers from a dead time of around 0.37 s, which is the time to observe a response to a step in the PWM steering input value. The analogue servo in one of the tractors has an actuation time delay of 0.2 seconds according to [6], the digital servo which is used in the other tractor has an actuation time delay of 0.09 seconds according to [7].

• The interaction between the TruckLab’s surface and the tires of the tractor result in tire slip.

Therefore even if the steering system can be accurately actuated, the desired path might not be followed due to tire slip.

• The steering system is not equipped with Ackermann steering, this is due to the steering rod connecting the front of the steering knuckles, this results in reverse Ackermann steering.

The wheel on the outside of the curve turns more than the wheel on the inside of the curve.

This results in additional resistance while cornering.

• The steering system has too many links, resulting in additional free play.

• The steered front wheels can move up and down unpredictably due to the leaf-spring suspension which connects the front axle with the chassis. This makes the direction of travel of the tractor more unpredictable.

Propulsion system

• The drive motor which propels the tractor has a low resolution, which results from the small PWM range (47 steps) which is available to control the drive motor. This low resolution results in oscillations in longitudinal velocity. Furthermore there is a motor delay of 0.6s.

• The longitudinal velocity of the tractor is controlled in open loop, because the tractor is not equipped with sensors to measure velocity. Therefore closed loop control of the longitudinal velocity of the tractor is not possible.

• The propulsion system suffers from a longitudinal delay of 0.52 s, which is the time to observe a response to a step in the PWM velocity input value.

• The tractor is not equipped with a brake system, therefore the tractor cannot be decelerated quickly enough. The only option is to decrease the velocity of the drive motor, this however results in stopping distances which are too long when decelerating from high forward velocities.

• A decrease in longitudinal velocity has been observed when the steering angle of the tractor increases. Therefore the prescribed longitudinal velocity cannot be maintained during a manoeuvre which involves cornering.

• The rear axle, which includes the differential, contains plastic parts which results in wear over time.

• The driven rear wheels can move up and down unpredictably due to the leaf-spring suspension which connects the rear axle with the chassis. This makes the velocity of the tractor more unpredictable.

• The drive time is limited by the battery capacity to approximately 30 minutes according to [5].

(18)

Olimex E407 micro-controller

• For development of the software on the Olimex control board there is a dependency on HANcoder to enable the use of Simulink models on the Olimex.

• The Olimex control board lacks voltage and current diagnostics, filtering and some protec- tions are needed to safely connect the steering servo and drive motor. The Olimex can otherwise be damaged by current peaks generated by the motor or the servo.

• The location of the Olimex control board in the semitrailer of one of the scaled tractors is a disadvantage, because some wires need to be routed from tractor to trailer.

1.2.2 Research objective

Based on the problem definition a research objective will now be formulated.

The research objective for this graduation project is to improve the steering system and the propulsion system and to address the problems formulated for the scaled tractor in general. The problems related to the Olimex E407 micro-controller and the possible replacement or extension of this board with additional hardware can be addressed in case it is necessary to improve the other subsystems.

The improvement which is gained in terms of accuracy and precision by these individual subsystems will be demonstrated by means of experiments. If possible within the time span of the project, experiments which test the entire closed loop system including the position feedback from the overhead camera system via the Wi-Fi based communication system are proposed.

1.3 Research plan

The research plan consists of the following items:

• Develop an improved steering system and propulsion system in order to solve the problems related to these systems described in section 1.2.1. The intended method is to:

– select/order/adapt/fabricate new steering system and drivetrain components on the basis of a thorough review of available parts. These components include steering servo, front axle, components enabling position feedback of the steering system, wheel bear- ings, tires, drive motor, components enabling velocity feedback of the propulsion system, rear axle, electronic speed controller and batteries.

– investigate the source of the longitudinal and lateral dead times and reduce it.

– remove unnecessary links in the steering system.

– improve the connection between the chassis and the steering system and the connection between the chassis and the rear axle by developing a more rigid connection.

– study literature on steering systems and propulsion systems used in scale models for research purposes.

The intended result is:

– A steering system and a propulsion system which:

∗ can be actuated with a higher resolution.

∗ have less movement with respect to the chassis.

– A steering system which has:

∗ an increased maximum steering angle.

∗ the possibility of measuring steering position.

(19)

∗ less free play.

∗ a reduced lateral dead time.

∗ less tire slip.

∗ less friction while cornering.

– And a propulsion system which has:

∗ a smaller motor delay.

∗ the possibility of measuring velocity of the tractor to enable closed loop velocity control.

∗ less longitudinal dead time.

∗ the ability to maintain a constant velocity during cornering.

∗ a longer lifetime due to reduced wear.

∗ increased drive time by higher battery capacity.

• Develop an improved chassis to create more space for necessary components and to obtain easier mounting of components on the chassis and reduced fragility of the tractor. The intended method is to:

– remove unnecessary components and keep the components which are needed.

– manufacture a new chassis out of metal, which allows for easy mounting of the necessary components. As the manufacturing will be done by Pieter van Hoof, my role will be to design the chassis and ensure that it meets the requirements.

The intended result is a simplified and more sturdy chassis, which offers more space to necessary components.

• Define, prepare and execute experiments to test the developed subsystems and the complete vehicle. If possible do experiments which test the entire closed loop system including the position feedback from the overhead camera system via the Wi-Fi based communication system. The experiments are done to show that the subsystems and closed loop system indeed have improved and meet the system requirements and to identify further points for improvement.

The intended method is to perform experiments with the improved tractor and analyze the results. These experiments should test whether a problem has been (partially) solved.

The experiments should deliver experimental results which compare the existing and new vehicles and should lead to the definition of standardized experiments which can be repeated as to indicate the possible further improvement of subsystems (and closed loop system) in the future.

1.4 Structure of the report

The structure of this report is as follows. Chapter 2 reviews literature published on the topic of remote controlled scaled vehicles which are used for research. Chapter 3 discusses the system requirements for the new tractor and the selection of components based on the requirements.

Chapter 4 gives a description of the new tractor. Chapter 5 discusses control of the tractor.

Chapter 6 discusses the experiments. In chapter 7 conclusions are drawn and recommendations for future work and improvements are given.

(20)

Chapter 2

Literature survey

2.1 Introduction

Remote controlled scaled vehicles are used by many researchers to do experiments with autonomous driving. Research has been done regarding the hardware which is used at these other facilities.

2.2 Remote controlled scaled vehicles for research

In [8] a low-cost RC toy car receives commands from a PC in order to follow a path. The car is shown in Figure 2.1. The Zigbee communication protocol and an ATmega 162 microcontroller are used in this car. An OptiTrack/Vicon system, which consists of 12 cameras, is used to get the positions and orientations of the car. It is stated that the Vicon system is more accurate as it can cover more area and has a larger range compared to the OptiTrack system. The need for robust control algorithms is described, as the RC car does not have precise control of its steering mechanism. As the battery discharges very quickly, the duty cycles which are used to steer the car need to be adjusted constantly in [8]. Also the car’s steering control does not respond correctly if the instructions are sent at a higher frequency. In [8] it is also observed that the surface friction has a large effect on the RC car’s motion. From [8] it can be learned that the use of toy cars, which are not build for a research application can give issues similar to the ones observed in TruckLab.

Figure 2.1: low-cost RC toy car [8]

(21)

CHAPTER 2. LITERATURE SURVEY

Car manufacturer Audi developed a 1:8 scale model vehicle specifically for the Audi Au- tonomous Driving Cup [10]. The scale model is shown in Figure 2.2. Students participate in this event to develop autonomous driving functions. This vehicle has two separate energy circuits.

One circuit for the ’mini-ITX board’ computer and one for the brushless motor. This brushless motor is controlled by a cruise control which gives the ability to drive forward, reverse and brake in a controlled manner, also at low speeds. The digital steering servo has integrated control, therefore a separate measurement of the steering angle is not necessary. Battery cell voltages can be monitored by an Arduino. The car is equipped with lidar, two mono video cameras, five ultrasonic sensors, wheel speed sensors and a 9-axis motion tracking sensor. The software used is ADTF(Automotive Data and Time-triggered Framework). Reference [10] shows the interest of car manufacturers in autonomous driving and gives inspiration for further development of the scaled tractor semitrailers used in the TruckLab.

Figure 2.2: scale model specifically build for AADC [10]

In [9] an autonomous RC car is discussed, which is the result of an interdisciplinary project at a university in Germany. It is stated that engineers always have to make a compromise between the power consumption and the available computing power. This trade off leads to heterogeneous computing systems which offer a beneficial power consumption. Learning how to integrate such systems into vehicles, is one of the aspects which can be investigated by means of a scaled RC car.

The result is a car that has a monocular camera, a lidar, a radar, differential GPS and an IMU.

The sensor data is communicated to the processing board by means of UART. The ZedBoard processing board consists of an Altera Cyclone 5 SOC with an FPGA and a dual core processor. The travelled distance of the car is obtained by motion sensors from the electric motor. The ultrasonic sensors are connected by means of an Arduino Uno. A separate TAPAS three phase inverter board is used to drive the motors. A raspberry Pi 3 is used to communicate to the outside world. The FPGA controls the steering servo motors. In total five types of communication protocols are used in the system. A schematic layout of the car is shown in Figure2.3. Reference [9] claims better energy consumption than the vehicles which are used in [10].

(22)

Figure 2.3: schematic layout of scaled car in [9]

An online project [12] describes a small robot, which uses an Arduino for processing. The robot is extended with a Raspberry Pi. The vehicle is shown in Figure2.4. It is stated that an Arduino is useful for interacting with sensors and motors. A Raspberry Pi on the other hand is useful to run computationally expensive algorithms like sensor fusion and image processing. Therefore the combination gives an inexpensive, but computationally powerful autonomous vehicle. The open source software library OpenCV is used for computer vision. Additionally ROS (robot operating system) is installed on the Raspberry Pi for development of robot applications.

(23)

Figure 2.4: scaled vehicle which combines a Raspberry Pi and an Arduino [12]

The onboard computers used in [8], [9], [10] and [12] have been compared to the Olimex onboard computer used in the existing tractor semitrailer of TruckLab. The computing power of an external PC as used in TruckLab and in [8] exceeds the computing power of the onboard computers in terms of memory and processor speeds. ’The mini-ITX-board’ as used in [10] has the most computing power of the onboard computers. The dimension of 17 cm x 17 cm is however too large to fit inside the tractor of TruckLab. All the other onboard computers would fit. Next the Raspberry Pi as used in [9] and [12] has the most computing power. A Raspberry Pi is a computer which is able to run multiple programs simultaneously. The other onboard computers in descending computing power can only run one program at a time: The Zedboard used in [9], The Olimex as used in the existing tractor, the Arduino used in [9] and [12] and lastly the Atmega 162 microcontroller as used in cite [8]. The tractors of Trucklab should be able to perform multiple tasks at once, such as processing sensor data, controlling actuators and detecting obstacles. There- fore a Raspberry Pi is a suitable option to use for the new tractor of Trucklab instead of the Olimex.

From [9] and [12] it can be concluded that there are different ways in which hardware can be combined in order to reach the desired result. However, the system architecture can get quite complex. The improvement in terms of energy efficiency in [9] compared to [10] seems to result in added complexity. From this it can be learned that it is not the case that an improvement in one area, results in an overall better system. It is therefore important to identify what the exact problems are in the TruckLab and which ones should be addressed in order to reach the desired result. It should be decided which parts can be preserved from the original tractor semitrailers, which were not developed for research purposes. New hardware needs to be selected taking into account the effects it has on the complete functioning of the TruckLab system. From [12] it can be seen that adding a single board computer like a Raspberry Pi to the tractors from TruckLab would enable the use of additional software tools. This might be useful to further extend the capabilities of the TruckLab.

2.3 Summary

In this chapter some remote controlled scaled vehicles used for research have been discussed.

The use of toy cars, which are not build for a research application can give issues similar to the ones observed in TruckLab. The vehicles discussed show the interest of car manufacturers in autonomous driving and give inspiration for further development of the scaled tractor semitrailers used in the TruckLab. A compromise between the power consumption and the available computing power needs to be made in a scaled vehicle. A Raspberry Pi is a suitable onboard computer to use for the new tractor of TruckLab.

(24)

Chapter 3

System requirements and component selection

3.1 Introduction

The problems found in the existing tractors have been categorized for the main subsystems. The most important subsystems are the steering system, the propulsion system and the computing/- electronic system including sensors. These subsystems are connected by a chassis. The tractor relies on software for control. To be able to improve the scaled tractors, other remote controlled scaled vehicles and robots as well as components have been researched.

Based on the outcome of this research and the identified problems, the requirements for our improved tractor have been determined. Based on the requirements, components have been selected and parts have been manufactured. In this chapter the requirements and the selection of components are discussed per subsystem.

3.2 Steering system

3.2.1 Steering system requirements

The requirements for a new steering system are:

• For accurate path following, slip of the steered tyres needs to be minimized.

• To reach the desired cornering radius of 0.49 m, the maximum steer angle needs to be increased. The maximum steer angle of a typical tractor is 49 degrees according to [2].

Under the assumption that the new scaled tractor has Ackermann geometry and adheres to the reference dimensions found in Appendix B, a maximum steer angle of 41 degrees is needed to reach the desired cornering radius.

• For repeatability of a steering manoeuvre, the free play in the steering system needs to be minimized.

• The steering system needs to be durable enough to sustain frequent use in TruckLab.

• The servo should be able to actuate the steering system under load and at standstill.

• The steering system needs to give positional feedback for monitoring and control purposes.

To see if the servo reaches the target position or to detect if something is interfering with it’s ability to get there.

(25)

CHAPTER 3. SYSTEM REQUIREMENTS AND COMPONENT SELECTION

• The steering system should respond quickly enough to control signals. The target is that the steering system should respond within 1 second to a control signal. 1 second is comparable to a quick reaction time of a driver.

• To be able to set the steering angle accurately, the servo should have a high resolution. The target is to have a resolution around 0.1^◦ in order to mimic the continuous steering output which is normally given by a driver. A target far below 0.1^◦ is not realistic, as the remaining free play in the steering system will most likely be larger.

• The steering servo should be compatible with the servo controller.

• The steering system needs to fit inside the tractor.

3.2.2 Component selection for the steering system

Based on the requirements for the steering system, components have been selected. The main components which need to be selected for the steering system are a front axle and a steering servo for actuation.

The number of available front axles to choose from is limited, because it should adhere to the scale of the tractor. A choice has been made between a metal or plastic front axle. A metal front axle has been selected for increased durability. The higher manufacturing precision should result in less free play. The selected axle has multiple mounting points to connect a steering link and the tie rod. This gives some adaptability to obtain the desired movement of the steered wheels.

The steered wheels should follow the Ackermann geometry such that tyre slip is limited.

There are many servo’s to choose from. Important considerations include whether the servo is digital or analog, material of the servo, power requirement, type of electric motor, response time, amount of backlash and dead band, encoder type, available rotation angle, programmable or not, feedback or not, available operating modes, connection to servo controller and finally available interfaces. An article explaining servo’s extensively is given by [15].

Servo’s used in RC vehicles mostly do not offer positional feedback. If they do, it is by means of an additional wire which is connected to the potentiometer inside the servo. A separate connection to a micro controller needs to be made in order to read the feedback signal. Most of the available RC servo’s with feedback have plastic reduction gears. Those result in low durability and backlash between gears. Most RC servo’s also have a limited rotation angle of 180 degrees maximum.

Digital servo’s generally have a smaller dead band compared to analog servo’s. Their positional accuracy is therefore often higher. Dead band of a servo describes the minimum change in control signal which results in a positional change of the servo. Digital servo’s are therefore more suited to prescribe the steering angle with a high accuracy. Digital servo’s can also achieve higher position refresh rates. Therefore the selected position is maintained better compared to most analog servo’s. Some digital servo’s can even be programmed.

As RC servo’s, also known as hobby servo’s, are not meeting the requirements, the use of servo motors has been explored. Servo motors give position and velocity feedback using a contactless encoder over 360 degrees, hobby servo’s use a potentiometer with a limited range. These servo motors are used often in robotics for their high positional accuracy. Dynamixel actuators are an example of such servo motors. Dynamixel is an actuator which integrates a DC motor, reduction gear, controller, driver and network capability into one module. Dynamixels are also programmable. To actuate the steering system a Dynamixel XM430-W210T has been chosen. An overview of this servo motor is given by [17]. The metal gears and casing of this specific model make it durable. The metal gears should also result in less backlash and therefore less free play.

The contactless absolute encoder provides positional and velocity feedback over 360 degrees. The resolution of 4096 steps per revolution means that the Dynamixel will not limit the precision with

(26)

which the steering system can be positioned. The limit on positional precision will be a result of remaining free play in the steering system as any remaining free play will most likely exceed the resolution of the Dynamixel of 0.088^◦/step. The stall torque of 2.7 Nm should be high enough to actuate the steering system. Because it exceeds the sufficient 2.15 Nm of the servo [6] used in the existing tractor. Dynamixel interfaces with multiple programming languages including MATLAB.

The dimensions of Dynamixel are suitable to fit inside the new tractor. A full overview of the selected Dynamixel servo motor is given by [14].

3.3 Propulsion system

3.3.1 Propulsion system requirements

The requirements for the new propulsion system are:

• The propulsion system should have enough torque to propel the tractor with the trailer, combined weight is around 10 kg.

• The propulsion system should have enough power to reach the desired top speed, around 1 km/h is sufficient for the manoeuvres done in TruckLab. At full scale this translates to approximately 13km/h, which exceeds the velocities observed during typical manoeuvres at distribution centers.

• The propulsion system should have a high enough resolution to vary the velocity smoothly as to be representative for a real tractor. The Assumption is that in a real tractor the velocity can be varied with a resolution of 1 km/h or 0.28 m/s. This number should be divided by the scale, which results in a target resolution of under 0.02 m/s for the scaled tractor.

• The propulsion system needs to give velocity feedback for monitoring and control purposes, to see if the target velocity is reached.

• The propulsion system should respond quickly enough to control signals. The target is that the propulsion system should respond within 1 second to a control signal, just as the steering system.

• The propulsion system should be able to maintain a constant target velocity without oscil- lation, both during straight line driving and cornering.

• The propulsion system should also be able to decelerate the tractor. A full scale tractor must be able to decelerate with at least 5.0 m/s²at a dry and level road according to Dutch legislation [28]. At scale this becomes a target deceleration of 0.38 m/s².

• The propulsion system should drive the wheels so that tyre slip is minimized.

• Free play in the propulsion system needs to be minimized for repeatability.

• The propulsion system needs to be durable enough to sustain frequent use in TruckLab.

• The propulsion system needs to fit inside the tractor.

• The drive motor of the propulsion system should be compatible with the electronic speed controller.

(27)

3.3.2 Component selection for the propulsion system

Based on the requirements for the propulsion system, components have been selected. One option is to create an improved propulsion system with a similar layout as on the old tractor. This layout consists of an electric motor typical for RC cars. This motor is controlled by an electronic speed controller which connects to a micro controller. In case of the old tractor the micro controller is the Olimex. For speed reduction this motor is connected to a gearbox. This gearbox connects to a prop-shaft. This prop-shaft connects to a mechanical differential in the rear axle. The rear axle holds the wheels.

An advantage of this layout is that the mechanical differential in the rear axle allows the individual rear wheels to rotate at the velocity necessary for cornering. A disadvantage of this layout is the indirect connection between the rear wheels and the motor. Each additional component used in this layout results in additional free play. A disadvantage of the mechanical differential occurs when one of the tyres looses traction. All power will be send to that particular wheel, which results in a spinning tyre. Therefore it can be beneficial to have control over the velocities of each rear wheel individually. The ability to control these velocities separately, together with an appropriate control strategy can limit tyre slip. Limiting tyre slip is beneficial for the predictability of the travelled path of the tractor. This has been an important reason to choose for a new layout with an electronic differential.

The electronic differential is created using two Dynamixels. Each Dynamixel directly propels one of the rear wheels. The same benefits which have lead to choosing for Dynamixel to actuate the steering system hold for the propulsion system. Additionally this new layout eliminates the need for a separate ESC, gearbox and prop-shaft, thereby reducing the amount of free play. Also this new layout saves valuable space, because of the compactness it fits between the rear wheels of the tractor.

The specified stall torque of two Dynamixels of type XM430-W210T should be sufficient to propel the tractor with the trailer. Taking into consideration the maximum RPM of the Dynamixels and the wheel diameter, the top speed requirement should be met. In velocity control mode the resolution which can be obtained is 0.229 rev/min or 0.001 m/s assuming direct drive of the scaled tyres. With a range of 1024 steps, this should be enough to meet the needs for TruckLab. Velocity feedback of the rear wheels is enabled by using Dynamixel. The internal PID control of Dynamixel should be able to maintain a constant target velocity. The new propulsion system layout is expected to be durable, because of the small amount of components used. Also the build quality and material of Dynamixel is expected to increase durability of the propulsion system. Another advantage of this choice is that the communication with the propulsion system is identical to that of the steering system.

3.4 Computing system

3.4.1 System requirements

The requirements for the new computing system/electronic system inside the tractor are:

• The system should have wireless connectivity for remote communication with the external computing system, thereby enabling remote control of the tractor. Wireless connectivity is also useful for remote software upgrades.

• The system should have sufficient operating time, which leads to a battery capacity requirement. Target is 1 hour of fully instrumented driving.

• The dependency on HANcoder limits the development of software on the Olimex board.

Therefore this dependency needs to be removed.

(28)

• The system should be located inside the tractor, so that driving without a trailer connected is possible.

• The system should have some voltage and current diagnostics, like battery monitoring.

• The system should have sufficient interfaces for connectivity to sensors, actuators and other computers. Expected interfaces are general purpose input output (GPIO) ports, USB, camera serial interface, Wi-Fi, Ethernet, PWM.

• For a course in the Automotive curriculum, operation of the tractor by means of a Raspberry Pi is preferred.

• The system should operate safely and reliable.

• The system should have a battery which is easy to access for charging or replacement.

• There should be an active online community for support during troubleshooting.

3.4.2 Component selection for the computing system

Based on the requirements for the computing system components have been selected. The main conclusion based on the problem definition regarding the computing system has been that the Olimex needs to be replaced. The dependency on HANcoder for further software development and the use of J-tag to setup the Olimex board has proven not to be ideal. Furthermore the direct connection of the steer servo and drive motor could damage the board. This damage may occur by current peaks initiated by the drive motor. A proper motor driver with filtering should therefore be used in between the control board and the actuators.

An option has been to design a custom board. On this board the necessary motor driver, desired I/O, diagnostics and filtering could be integrated. This option however would pose time issues for this project. A preference has been expressed to replace the Olimex with a Raspberry Pi for an Automotive course given at university. There are some benefits of using a Raspberry Pi.

Some Raspberry Pi models have integrated Wi-Fi, this would remove the need for a separate Wi-Fi dongle as is used currently for wireless connectivity. The Raspberry Pi is also widely supported online. A downside of a Raspberry Pi is that it lacks analog I/O. Therefore additional boards need to be used to obtain analog I/O, in the case analog sensors or actuators are used. This can for instance be done by means of an Arduino as in [12].

During the search for actuators for the propulsion and steering system the conclusion has been drawn that the use of Dynamixel is preferred. Because of this decision the possibility to connect Dynamixel servo motors to a Raspberry Pi has been explored. A unicycle robot named

”TurtleBot3 Waffle Pi” from the company Robotis combines Raspberry Pi with two Dynamixels for actuation. An overview of the robot is given by [16] and shown in Figure 3.1. At [16] the assembly manual of Turtlebot3 is also available. An extra ”OpenCR1.0” board is connected to the Raspberry Pi via USB. This board is an open source robot controller specifically designed to control Dynamixel servo motors. An overview is given by [18]. This board supplies power to the robot by means of a LI-PO battery or a switched mode power supply (SMPS). This particular SMPS is an AC to DC converter used to either power the robot from the wall socket or to charge the batteries. It is particularly convenient that the battery can be disconnected for charging, while power is still supplied to the robot via the SMPS. Therefore the robot can maintain online, whilst changing or charging the battery.

(29)

Figure 3.1: overview of Turtlebot3 Waffle Pi [16]

Because TurtleBot3 is modular and can be modified, this particular robot has been researched further. The main question has been to see whether this robot could be converted to a tractor for TruckLab. Some requirements for the electronic system still need to be investigated.

The battery capacity of the new tractor should allow 1 hour of fully instrumented driving. Speci- fied operating time for Turtlebot3 is 2 hours with the standard battery. The new tractor including trailer has an expected weight of 10 kg. Turtlebot3 weighs only 1.8 kg. Therefore the operating time is expected to reduce, when using the same battery for the new tractor. At low battery voltage a buzzer will sound. While the battery type used is fairly standard, a higher capacity battery can be selected in case necessary.

The selected electronic components should fit inside the tractor. In case of the electronic components of Turtlebot3, this has been verified to fit by doing measurements and drawing possible layouts for the new tractor. The interfaces of Turtlebot3 provided by the Raspberry Pi and the OpenCR board are sufficient to drive the Dynamixel actuators and the standard lidar and camera.

The digital inputs of the Raspberry Pi should be sufficient to connect additional digital sensors if necessary. The possibility to extend the OpenCR board by connecting an Arduino Uno would enable the connection of additional analog sensors as well. The standard software used for Turtle- bot3 allows monitoring of diagnostic messages. This diagnostics includes battery, actuator and sensor status. As Turtlebot3 is a commercially available robot it should be safe, as long as it is used with care. The electronics system does not need to be modified heavily for possible use in the new tractor. As commercial products are often tested before being sold, expectancy is that Turtlebot3 will offer some reliability as well.

The electronic system/computing system of the Turtlebot3 can meet the requirements with some additions or modifications. It will therefore be used as a basis for the electronics of the new tractor.

3.5 Sensors

3.5.1 Sensor requirements

The requirements for the new sensors are:

• The sensor package on the new tractor should be extendable.

(30)

• The sensor package should have the capability to measure the position and orientation of the tractor.

• The sensor package should have the capability to measure the velocity of the tractor.

• The sensor package should have the capability to measure the position of the steering system.

Using this position the front wheel steering angle δ can be obtained.

• The sensors should be compatible with the electronics system/computing system to be able to process the sensor data.

• It should be possible to extend the sensor package for obstacle detection.

3.5.2 Sensor selection

Based on the requirements sensors have been selected. As concluded, in section3.4.2, Turtlebot3 offers enough extensibility for additional sensors. The sensors included in Turtlebot3 will form a basis. The available open source software packages for various sensors should make adding a sensor to Turtlebot3 easy. Turtlebot3 comes with a lidar and a Raspberry Pi camera. The lidar takes range measurements over 360 degrees in a 2 dimensional plane. It can be used for obstacle detection and mapping of the environment. The Raspberry Pi camera can be used to implement camera vision. The selected Dynamixels will be used to measure the velocities of the rear wheels and the position of the steering system. The encoders of the two Dynamixels which actuate the rear wheels can also be used to calculate the position of the tractor relative to a starting point by using odometry. Together with the lidar this gives SLAM capability. The openCR board of Turtlebot3 has an IMU integrated. The IMU consists of a 3-axis gyroscope, 3-axis magnetometer and 3-axis accelerometer. As the sensor package of Turtlebot3 is already quite complete, the decision is to extend it when necessary.

3.6 Software

3.6.1 Software requirements

The requirements for the new software are:

• The software should provide the capabilities needed for TruckLab. These capabilities should include processing sensor data, local low level control of actuators inside the tractor, central high level control of the tractor for autonomous manoeuvres on an external PC.

• Controllers used to perform autonomous manoeuvres should be developed in MATLAB/Simulink, as this software is widely used across university.

• The software should be adaptable to allow new capabilities for future development of the tractor.

• The software should be supported by the computing system.

• There should be online support for the software.

• The components should be able to interact with each other in case various software is used.

.

(31)

3.6.2 Software selection

Based on the requirements for the software, software has been selected. The use of MAT- LAB/Simulink has been obvious, as this software is widely used across university for control system design. Users of the TruckLab can therefore likely benefit from previous experience with MATLAB/Simulink. Therefore the high level control of the tractor for autonomous manoeuvres will be done in MATLAB/Simulink.

Turtlebot3 uses ROS(Robot Operating System). The conversion of Turtlebot3 into a new tractor for TruckLab requires ROS to be capable as well. Therefore the possibilities of ROS have been investigated. ROS can interact with MATLAB and Simulink using the ROS Toolbox and the Robotics System Toolbox available for MATLAB and Simulink. The ’ROS Subscribe’ block in Simulink can receive messages from the ROS network. The ’ROS Publish’ block in Simulink can send messages to the ROS network. These blocks provide the coupling between Simulink and ROS. There are several benefits of using ROS alongside MATLAB and Simulink. ROS provides open source software packages for robotic components and robotic applications. This reduces the need to write software from scratch. Furthermore ROS includes various useful robotic tools. There are several tools for visualizing sensor data and a tool to see the message communication between components of the robot. As ROS is open source, all the software is adaptable. Just as MATLAB and Simulink, ROS is well documented. This documentation should make learning the basics of ROS easy. ROS programming in combination with use of Turtlebot3 is explained in [23].

The firmware of Turtlebot3 is loaded on the openCR board. This firmware communicates via USB with ROS on the Raspberry Pi. This firmware is also adaptable. This adaptability is necessary to convert the Turtlebot3 into a tractor. The new tractor will use an added Dynamixel for steering. To make this Dynamixel operational the firmware needs to be adapted.

The use of MATLAB and Simulink in combination with ROS and the Turtlebot3 firmware does not pose any foreseen issues. Therefore the conclusion is that Turtlebot3 is suited to form the basis for the new tractor.

3.7 Chassis

3.7.1 Chassis requirements

The requirements for the new chassis are:

• The mass of the chassis should be sufficiently low. The Dynamixels should be able to propel the combined weight of the tractor and semitrailer(s). Therefore the goal is for the new tractor to weigh less than 4 kg. This is similar to the weight of the existing tractor. The total weight of the Turtlebot3 is 1.8 kg. Taking into account the weight of other components, like the cabin, wheels, tyres and front axle, the target weight for the chassis is 1 kg or less.

• The chassis should reflect the dimensions of an actual tractor, but scaled down with a factor of 13.3.

• The chassis should be stiff enough as to not impair the desired accuracy of the tractor.

• The chassis should have space for the components to be mounted at the location and with the orientation needed to fulfill their function.

• The chassis should be simple to assemble and disassemble.

(32)

3.7.2 Chassis design

An option which has been researched is to adapt the chassis of the existing tractor. However, it has been concluded that this is not the favourable option. The old chassis needs too much modification to properly attach the new components. Furthermore it is part of the old toy tractor, which requires a lot of assembly. Using the old chassis as a basis for the new tractor is not efficient.

Each new tractor would then require the use of a tractor similar to the old one for parts. Separate chassis parts are available at very high prices. Therefore it has been decided to design a new chassis.

It has been considered to reconfigure the modular plates of Turtlebot into a tractor chassis. The fixed dimension of these plates of 7 cm x 14 cm however is a limitation. Therefore it has been decided to develop a new chassis out of metal. The chassis dimensions can then be made according to the requirements. Some modular plates of Turtlebot can still be used in combination with the metal chassis. The modular plates have plenty of holes to easily connect and relocate electrical components. The new chassis will be manufactured out of aluminum. The low density of aluminum compared to other metals should give the chassis the desired target weight.

3.8 Summary

In this chapter the requirements for our improved tractor have been discussed per subsystem.

Based on the requirements, components have been selected and parts have been manufactured.

As RC servo’s, also known as hobby servo’s, are not meeting the requirements, the use of servo motors has been explored for the steering system and the propulsion system. In the new propulsion system the rear wheels are driven individually by two separate motors. The majority of the selected components such as the Dynamixel servo motors and the Raspberry Pi originate from a unicycle robot named ”Turtlebot3 Waffle Pi”. This robot will be used as the basis for the new tractor.

This robot comes with an already quite complete sensor package. The ROS software used for this robot will interact with MATLAB/Simulink to control the new tractor. It has been decided to design a new chassis instead of modifying the old chassis.

(33)

(34)

Chapter 4

System description of the tractor

4.1 Introduction

Components have been selected based on the system requirements. Robotic components have been combined with RC components and in house developed components. Robot Operating System (ROS) software is used for extra software tools and interacts with MATLAB/Simulink to control the new tractor.

This chapter gives first an overview of the closed loop control system used in the TruckLab.

The scaled tractor semitrailers are a part of this system. Next an overview of the locations of the components inside the new tractor is given. Thereafter the component integration is discussed on the basis of design choices related to the connection of components.

4.2 Closed loop control system overview

An overview of the closed loop control system of TruckLab is given in Figure4.1. The communication between the main components are:

Model tractor communication to overhead camera system

The ArUco markers on the tractor and trailer are observed by the camera system. Together with the stationary ArUco markers on the floor the pose of the tractor and trailer can be determined.

Overhead camera system communication to tractor control

The camera system outputs the ArUco Marker positions and orientations to tractor control. This pose information is received in MATLAB/Simulink on the external PC with Linux.

tractor control communication to model tractor

Tractor control consists of a MATLAB/Simulink based control system on the external PC with Linux, which outputs the required control signals for steering and propulsion to ROS. ROS on the external PC with Linux communicates these commands via Wi-Fi to the Raspberry Pi 3 with ROS in the model tractor. The model tractor has a bluetooth module which can be used to receive manual commands from a bluetooth remote controller instead.

Model tractor communication to tractor control

The model tractor communicates Raspberry Pi Camera data, Laser Distance Sensor data, Dy- namixel servo feedback and IMU data to tractor control. The Dynamixel servo feedback consists of position, velocity, current, realtime tick, trajectory, temperature and input voltage information.

IMU data consists of 3axis gyroscope, 3axis accelerometer and 3axis magnetometer information.

This set of feedback information is useful for tractor control and monitoring of the tractor.

(35)

CHAPTER 4. SYSTEM DESCRIPTION OF THE TRACTOR

Figure 4.1: Overview of the closed loop control system of TruckLab

(36)

4.3 Tractor system overview

An overview of the new tractor is given in Figure4.2. The chassis consists of two aluminum plates with spacers in between. The other components are attached to the chassis. The openCR board, the battery and the bluetooth module are mounted on waffle plates in between the front and rear wheels. The steering system is located between the front wheels. The propulsion system is located between the rear wheels. On the top chassis plate the fifth wheel is mounted. Above the front wheels a stack of waffle plates is attached to the chassis. This stack contains the Raspberry Pi, the Raspberry Pi camera, USB2LDS and the Lidar.

Figure 4.2: Overview of the new tractor with the cabin removed

4.4 Component integration to form the new tractor

In this section the integration of Turtlebot components, RC components and in house developed components is discussed. The integrated components form the new tractor. The location and orientation of some components is explained. Before the components have been ordered several 2D and 3D CAD drawings in NX Siemens have been made to ensure the components would fit.

Examples of these drawings are shown in Appendix D.

4.4.1 Steering system

The new steering system is shown in Figure4.3. The main design question for the steering system has been how the Dynamixel should be connected to the front wheels. The orientation and position of the Dynamixel servo with respect to the front axle influences the relation between servo position and steering angle. This is further influenced by the length of the steering link and the connection points of the steering link. The front axle has multiple connection points to attach the steering link.

The other end of a steering link is usually attached to the servo by means of a lever, known as the servo horn. The position of the connection point of the steering link on the servo horn influences the ratio between servo rotation and steer angle. This ratio influences the resolution with which the steer angle can be set. The resolution increases if the servo rotates over a larger angle. Of further importance is that the steering mechanism should be able to move without coming into contact with other components of the tractor. Furthermore the number of steering joints of a steering mechanism influences the amount of free play in the system. The manufacture-ability and the ease of mounting the steering system to the tractor chassis should also be considered.

(37)

Figure 4.3: The new steering system

Several options for the connection between the Dynamixel servo motor and front wheels that have been considered are:

1. Mount the steer actuator in such a way that its axis of rotation is not parallel to the inclined axis of the kingpin. Then connect the steer actuator with the steering knuckle or with another connection point via one or multiple links.

2. Mount the steer actuator in such a way that its axis of rotation is parallel to the inclined axis of the kingpin. Then connect the steer actuator with the steering knuckle or with another connection point via one or multiple links.

3. Mount the steer actuator in such a way that its axis of rotation is parallel to and coincides with the inclined axis of the kingpin. Then connect the steer actuator with the steering knuckle via one or multiple links.

Options which fall into category 1, result in one or multiple links which need to translate in three directions, up-down, left-right, forwards-backwards. Therefore it is difficult to analyze be- havior of such a mechanism and the influence of design parameters as for instance length of links.

This can be done with a multibody model of the mechanism. The relation between the position of the actuator and the position of the steering knuckle will be nonlinear. Mounting of the servo is probably easiest choosing an option out of category 1. Steering accuracy is probably degraded by the use of several joints necessary to make the required connection. Joints will result in free play.

Options which fall into category 2, result in one or multiple links which need to move in a plane.

This plane is perpendicular to the inclined axis of the kingpin. A mechanism which moves inside this plane is probably somewhat easier to analyze compared to options out of category 1. The relation between the rotation of the actuator and the rotation of the steering knuckle will still be nonlinear. Mounting of the servo is a bit more difficult, because it needs to be inclined.

Options which fall into category 3, result in a link which only needs to rotate around the co- inciding axis of the kingpin and the steering actuator. This results in a linear relation between the rotation of the actuator and the rotation of the steering knuckle. As the relation is linear, the servo angle is equal to the steering knuckle angle, analysis of this mechanism is simple. Steering accuracy is highest as the servo, link and steering knuckle can rotate as a single body. Mounting of

Eindhoven University of Technology MASTER Development and testing of a scaled tractor for use in the TruckLab Smeulders, G.A.M.