Motion planning for an autonomous terrestrial ehicle in static environments
Andries Jacobus Bester · SUNScholar (Stellenbosch University) · 2018
ENGLISH ABSTRACT: Ground or land-based vehicles capable of navigating autonomously in various environments are rapidly becoming more popular and have multiple advantages and uses, from a Mars rover to autonomous agricultural vehicles or search-and-rescue robots. Vehicle manufacturers are also racing to be the first to deliver a completely self-driving car, with technological advances already facilitating features like self-parking, predictive braking, adaptive cruise control and traffic sign recognition. This project aims to contribute to autonomous navigation research by improving on motion planning techniques for terrestrial vehicles. The work done includes the modelling and control of a quad bike with actuators installed to facilitate autonomous control of the steering, throttle and brakes of the vehicle. With the simulation model of the vehicle, the commanded behaviour of the vehicle could be quantised in sets of movements to produce replicable manoeuvres. The path planning algorithm developed in this project then searches in a map of the environment and uses these manoeuvres to build a path for the vehicle, which adheres to the kinematic and dynamic constraints of the vehicle. To supplement the modular design of the path planning algorithm, a simplified conflict detection algorithm and map representation were created to allow the path planner to be developed without knowledge of the map or the representation of obstacles. The results obtained from the field tests indicate that the path planning algorithm could find a path that became more optimal the longer the algorithm was allowed to search, and that the vehicle could follow the path generated by the planner. The functionality of the path planner was demonstrated without the ability to do replanning in real time, and is therefore advised that further research should include the real-time application and optimisation of the proposed algorithm.