Abstract:
The Ilon Mecanum wheel is one of the practical omnidirectional wheel designs in industry but it consumes excessive energy due to its motion inefficiency. Omnidirectional autonomous mobile robots without any nonholonomic constraints can reach the same goal pose from the same start pose via too many feasible trajectories, each consuming different amounts of energy. It is both economically beneficial and academically interesting to optimize the omnidirectional trajectories of holonomic Mecanum autonomous mobile robots for energy minimization. The aim of this PhD research was to first develop a heavy-duty omnidirectional Mecanum mobile robot that can perform fully functional autonomous navigation for research purposes, and then to realize energy-efficient autonomous navigation on the designed robot via both online and offline optimal motion planning research studies. A heavy-duty omnidirectional Mecanum robot platform was designed and developed from the beginning. The robotic control system architecture that allows the robot to perform autonomous navigation was realized by fusing an industrial automation control system and open-source autonomous navigation technologies. Then an experimentally validated novel energy consumption model of the four-wheel omnidirectional Mecanum mobile robot was proposed based on a comprehensive understanding of the kinematics, dynamics and energy flow of the robot. In order to realize energy-efficient autonomous navigation, both online and offline optimal planners were presented in this research. The proposed online planner extended the Dynamic Window Approach (DWA) by means of a new energy-optimal omnidirectional velocity search technique for the purposes of optimizing motional power consumption and reducing energy consumption for autonomous navigation. The offline planner proposed a method of trajectory generation that interpolates the task-space trajectory of the omnidirectional robot in polynomial spline functions and then searches for the energy-optimal trajectories using genetic algorithms. Both proposed online and offline planning methods were experimentally validated.