Adaptive Lateral Control of Autonomous Trucks

Abstract

Autonomous heavy vehicle control systems depend on a suite of parameters, both vehicular and environmental, and their performance is directly related to these parameter values. These parameters often include variables such as the total mass of the loaded vehicle, lateral and longitudinal tire cornering coefficients, wheelbase geometry, load distribution, etc. In theory, these parameters are often considered constant and known. However, in practice, this might not always be the case, especially in dynamic operating environments of a quarry or mine. Therefore, to ensure robust and reliable performance, the controller must be adaptive, updating itself online to continue operating efficiently. This thesis focuses on the investigation, development, and implementation of different types of adaptive lateral controllers for autonomous haulage trucks, whose parameters initially deviate from the true value and are subject to change during operation. The study focuses on evaluating both direct and indirect approaches to adaptive control. The direct adaptive controller uses a Lyapunov-based approach towards adaptation, meanwhile the indirect controller uses Kalman filter-based estimation methods to estimate unknown system parameters. Overall, the findings suggest that the indirect adaptive control strategies are more suitable for autonomous haulage applications. The integration of Kalman filterbased parameter estimation with model based controllers like LQR resulted in consistent performance and robustness, and better handling of system uncertainties compared to direct adaptive approaches.