An efficient neural network method for real-time motion planning with safety consideration

An efficient neural network method is proposed for real-time motion plannin g of a mobile robot or a multi-joint robot manipulator with safety consider ation in a nonstationary environment. The optimal robot motion is planned t hrough the dynamic neural activity landscape of the biologically inspired n eural network without any prior knowledge of the dynamic environment and wi thout any learning procedures. The planned real-time "comfortable" path doe s not suffer from the "too close" or the "too far" problems. The proposed m odel is stable, computationally efficient, and not very sensitive to parame ter variations. The effectiveness and the efficiency are demonstrated throu gh simulation and comparison studies. (C) 2000 Elsevier Science B.V. All ri ghts reserved.