Real-time optimized obstacle avoidance for robotic vehicles: Indoor experiments

Recently, mobile indoor robots have been intensively studied. When such mobile robots are operated, it is necessary to generate their trajectory automatically. In this study, a system which automatically generates the target trajectory avoiding obstacles in the indoor environment by real-time optimization is developed. Conventional artificial potential field method directly generates force which leads vehicles, however, it sometimes cause steep change of moving direction. To resolve this problem, we generate an optimal obstacle avoidance trajectory indirectly by real-time optimization based on modified artificial potential field. To deal with the tracking error and its convergence, we introduce time-state control form and transform this problem into a tracking control problem. The trajectory is represented by a reduced order polynomial function to minimize the computational cost. Experiments using omnidirectional vehicle verify the advantage of this method.