PSEUDO-SPECTRAL METHODS BASED REAL-TIME OPTIMAL PATH PLANNING FOR UNMANNED GROUND VEHICLES

Abstract

Real-time optimal trajectory design and tracking for autonomous ground vehicles are maturing technologies with the potential to advance mobility by enhancing time and energy efficiency in application such as indoor surveillance robots or planetary exploration rovers. Pseudo-spectral methods based trajectory generation framework provides the desired trajectory which minimizes a prescribed objective function (i.e. minimum time, acceleration, and energy) while satisfying kinodynamics and various types of constraints (i.e. obstacle avoidance and smooth turning at waypoint transitions). In this thesis cyber-physical system architecture is used for the communication between rover-vehicle and the ground station. By using optimal state and control vector from trajectory generation module and by obtaining the state feedback values from the cyber-physical system architecture, a backstepping based controller provides commanded control values to complete the trajectory. Combination of novel optimal trajectory framework (Guidance), modified backstepping controller (control) and cyber-physical system architecture makes the complete guidance navigation and control system. This thesis work elaborates, the efficacy of the overall approach by performing several experimental test runs carried out with the rover vehicle equipped with GPS, compass, and wheel encoders.