Event-potential method for mobile robot motion planning and control

Abstract

Autonomous mobile robots, particularly wheeled robots, have found extensive applications across numerous fields. Motion planning and control are pivotal to a robot's autonomy and are essential for completing any task. Dealing with uncertainties and unexpected events in the environment is a general challenge for mobile robot motion planning and control. Considering that the actual environment is often highly dynamic, with considerable uncertainty and unpredictability, motion planning and control methods need to handle unexpected events and uncertainties in the environment, such as dynamic obstacles. Existing methods usually address this problem in a decoupled way; that is, every time an unexpected event occurs, the sensor first perceives new environmental information, the planner then re-plans a new feasible plan, which is to be executed by the controller. The robot frequently undergoes this process in the dynamic environment to cope with the challenges of the dynamic environment. This thesis proposes the event-potential method, which uses event reference based on real-time sensor measurements to replace the time reference, thereby making motion planning and control into a real-time closed-loop process. Therefore, robots can locally modify the pre-planned trajectories in real time. A modified potential field method is used to locally modify the desired velocity on the pre-planned trajectory for obstacle avoidance while maintaining the trajectory tracking. The robot can always find the event reference on the pre-planned trajectory, thereby eliminating the need to re-plan even after unexpected events. This method enables the robot to respond to environmental changes more efficiently and swiftly, removing the inefficient, time-consuming, and computationally expensive re-planning process.

Additionally, the event-potential method also proposes a framework for motion planning and control of mobile robots. The motion planning method employs graph optimization method to tackle non-linear multi-objective optimization problems for trajectory planning, based on the path planned by the A* algorithm. The objective functions considered in the optimization process include total time consumption and total travel distance. Constraints from robot kinematics and dynamics, waypoint constraints, obstacle constraints, and others, are also considered in the trajectory planning process. The event-potential method is implemented on a 3-wheeled omnidirectional mobile robot (3WOMR). Consequently, the motion control method introduces the kinematic and dynamic models of the 3WOMR and describes the form of the corresponding globally asymptotically stable control law. This method achieves better results in navigation experiments in dynamic environments when facing unexpected events compared to existing algorithms like the dynamic window approach (DWA) and timed elastic band (TEB). Finally, this framework can also be applied to a variety of robots and tasks, demonstrating strong scalability.