Hazard avoidance auto control system for a robotic vehicle

Abstract

It is estimated that 1.2 million people die yearly by traffic accidents (WHO, 2012), driver error being the sole or contributing factor in 60%, 95% of accidents respectively (S.J. Anderson, Date N.A.). Unlike humans, computers are not vulnerable to stress, fatigue or other conditions. Reactive braking technologies must detect a problem in order to act, whilst fully autonomous driving results in the driver becoming passive/lethargic, unable to act correctly in unforeseen situations (Charette, Automated to Death, 2009). The aims of this thesis are to design a control system for a physical robot/vehicle, which autonomously avoids hazards including avoiding collisions and rollover prevention, one of the deadliest traffic accidents. Automobiles are of primary concern, but the system is also viable for other transport means and mobile robots. TCAS and Envelope Protection used in Aviation, are of interest as the pilot/driver still retains control given his/her inputs conform to the controller set limits, and is free to avoid collisions in advance, or turn at slower velocities than controller values. The system should continuously monitor the vehicle, being in a position for both preventive and reactive action. A robot was constructed and tested in physical environments for realism. A pipelined and multithreaded Fuzzy and Crisp behavioral architecture has been realized exhibiting modularity, maintainability, efficiency, simplicity and scalability. The system provided a robust performance in face of variable communication delays, inaccurate sensing, crosstalk, and deliberately hazardous user-input, adapting to unseen and varying environments, with seamless transition between driver and machine control. The use of non-real-time operating system and programming language further demonstrates the robustness of the algorithms developed. Controller performance was consistent despite random/variable parameters such as wheel slippage, castor side effects (Roland Siegwart, 2011), and imperfections in the drive mechanism (Carmel Gafa', 2008) which can all deviate the robot from intended trajectories.