Development and Comparison of Intelligent and Deterministic Control and Path Planning Methods for Articulated Steering Mobile Robots
Abstract
Articulated steering mobile robots and vehicles have a power revolving joint in the center, allowing them to make sharp turns and perform complex maneuvers in different applications. These Articulated Steering Machines (ASMs) allow smaller Turning Radius compared to the car-like or tractor-trailer structures. This flexibility, in the meantime, creates challenges in control, path planning, and docking maneuver, which requires reaching the dock in a certain direction within a minimal displacement error. This thesis explores autonomous control challenges faced by these robotic systems, including maneuverability, stability, singularity, disturbances, and mechanical constraints, and suggests approaches to address these challenges.
First, a hybrid feedback controller is proposed which covers all plane quadrants, robot poses, and backward moves. This technique is sensitive to control parameters and may exhibit different behavior for each initial robot pose. To address this issue, a genetic algorithm (GA) is used to find the best control parameters for any docking scenarios with some limitations. Next, a nonlinear model predictive controller (NMPC) is proposed with a waypoint function. This controller optimizes the robot's path at every interval and uses an initial guess with waypoints to avoid docking walls. The proposed NMPC can adjust to varying docking module poses but takes a long computation time.
Another proposed method is called Bézier-based nonlinear controller which incorporates a path planning algorithm based on Bézier curves and a nonlinear PID tracking controller. The algorithm uses a derived turning radius formula, avoids obstacles, and creates a smooth path. The nonlinear controller follows this online path to reach the current docking pose.
Disturbance effect evaluations show that the hybrid controller and GA trained controller can handle fair levels of disturbances, while the NMPC and Bézier algorithm successfully manage high levels of unmeasured disturbances. Comparisons between the proposed methods as well as existing algorithms demonstrate the advantages of the Bézier-based nonlinear controller in terms of accuracy, path length, computation time, and robustness to disturbances.