Path Planning and Control in an Autonomous Formula Student Vehicle - Adam Slomoi 2018

Monash Motorsport Final Year Thesis Collection

The Final Year Thesis, is a technical engineering assignment undertaken by students of Monash University. Monash Motorsport team members often choose to conduct this assignment in conjunction with the team.

These theses have been the cornerstone for much of the team’s success. The purpose of the team releasing the Monash Motorsport Final Year Thesis Collection is to share knowledge and foster progress in the Formula Student and Formula-SAE community.

We ask that you please do not contact the authors or supervisors directly, instead for any related questions please email info@monashmotorsport.com

Summary

This report details the design of a path planning and control algorithm for a formula student vehicle. The algorithm first discovers the outline of a track. It does so by linking cones to form two track sides and implementing Model Predictive Control to drive at a slow speed along the centreline of the track. This is run continuously until the track has been discovered in its entirety, at which point the racing line is optimised on-line. The racing line is computed by optimising the positions and speeds of the vehicle at points along the track using the Levenberg-Marquardt Algorithm. This is enabled by the design of a custom built Automatic Differentiation class in C++. The algorithm is successful in its performance however is limited primarily by the computational speed, which increases exponentially with an increase in the number of points considered along the track.

Introduction

This project aims to design a Path Planning system for an autonomous Formula SAE vehicle. This algorithm provided a good foundation to build the updated system that was used to demonstrate the car in 2019. The system must be safe, reliable and compliant with Formula SAE rules and regulations. Path planning has an integral role in an autonomous vehicle. This involves the development of a software algorithm which builds a dynamic trajectory for the car to follow. This is a crucial step between the car perceiving its environment (via sensors) and the physical actuation of the car’s controls.

This project aims to develop such an algorithm which will allow Monash Motorsport to successfully complete all dynamic disciplines. To allow knowledge transfer to take place year-on-year and for the team to continually improve upon the software design, an emphasis will be placed on well-documented material and justifications of design choices.

This report showcases the theory behind the design of the path planning algorithm for all dynamic events. Given that the Trackdrive event constitutes the majority of dynamic points, it is the focus of the development. Additionally, given that the track for this event is unknown, it is presumed to be an encompassing solution; if path planning is developed to complete the Trackdrive event successfully, then with minor tweaking it should also be able to satisfactorily plan a trajectory for the other two dynamic events. In order to achieve this, the notion of ‘path planning’ is broken down into three factors:

1. Discovery Lap

2. Optimal Racing Line

3. Control

This project details the development of the first two factors listed above. Rudimentary Control was introduced to facilitate the development of the Discovery Lap and Optimal Racing Line, however its development is beyond the scope of this project.

Conclusion

The goals of this project were achieved satisfactorily. The discovery lap algorithm is able to continuously plan a path and follow it, enabling the vehicle to drive and learn a racetrack without any prior knowledge. The resultant cones are used to compute an optimal racing line around the track on-line. The racing line optimisation is successful and is able to be easily tuned to factor in different considerations, such as limiting speed. The computation time relies heavily on the size of the track being considered, and further work needs to go into optimising the code to hasten the execution time. Overall, the path planning is at a promising stage of enabling Monash Motorsport in achieving its goal of completing all dynamic events at a formula student driverless competition.

Afterword to 2021

Since Adam’s FYP was published, many improvements have been made to the methodology behind path planning. This particular project is still an instrumental basis of current designs; however, it is important to note what updates have been made. The autonomous department improved upon Adam’s discovery lap algorithm by using Delaunay triangulation to define neighbouring cones, achieving improved robustness even without detected cone colours. These refinements allowed the car to corner successfully in sharper and more complex corners, reducing lap times as the car can be pushed to attain greater speeds with more confidence. Improvements were made to the Model Predictive Control (MPC) and optimising the motion controller for each mission.