Trajectory optimization of a mecanum-wheeled-legged hybrid robot for obstacle avoidance

Abstract

Robots are expected to help humans in various environments such as disaster areas. For robots to perform their work in any environment, algorithms are needed that can cope with a given environment. However, such algorithms easily fail due to the complexity of robot dynamics and kinematics, and environment. For example, sample-based methods suffer from the curse of dimensionality and optimization-based easily fails due to poor initial guess. This paper presents a trajectory optimization framework of a mecanum-wheeled-legged hybrid robot that can cope with obstacles. The research proposes a hierarchical process to generate a trajectory using (a) sampling-based method and (b) optimization-based method. The sampling-based stage generates a trajectory of the robot body and leg joints, which speeds up the optimization’s convergence rate and enhances the success rate of optimization. The optimization-based stage considers all constraints including the mecanum wheel non-holonomic constraint and passive wheel constraint and generates a trajectory of the robot body and leg joints. The result shows that the combining sampling-based method and optimization method can generate trajectory even in complex environments such as step and stair. The dynamic simulator and real hardware system are used to verify that the generated trajectory is feasible.