Evolutionary programming-based univector field navigation methodfor past mobile robots

摘要

Most of navigation techniques with obstacle avoidance do notnconsider the robot orientation at the target position. These techniquesndeal with the robot position only and are independent of its orientationnand velocity. To solve these problems this paper proposes a novelnunivector field method for fast mobile robot navigation which introducesna normalized two dimensional vector field. The method provides fastnmoving robots with the desired posture at the target position andnobstacle avoidance. To obtain the sub-optimal vector field, a functionnapproximator is used and trained by evolutionary programming. Two kindsnof vector fields are trained, one for the final posture acquisition andnthe other for obstacle avoidance. Computer simulations and realnexperiments are carried out for a fast moving mobile robot tondemonstrate the effectiveness of the proposed scheme