In Optimal Stochastic Path Planning (OSPP) of industrial robots the problem can be represented by a variational problem under stochastic disturbances. Using deterministic substitute problems obtained from stochastic optimization theory, the corresponding deterministic substitute variational problem can be solved approximatively via mathematical programming techniques. However, due to the computationally expensive algorithms involved, these techniques can be used for off-line calculations only. Hence, neural networks are trained to solve the path-planning problem in real-time and to provide suitable solutions for on-line applications.
Dipl. Math. Andreas Aurnhammer, Aero-Space Engineering and Technology, Federal Armed Forces University, 85577 Neubiberg/Munich, Germany
Prof. Dr. Kurt Marti, Aero-Space Engineering and Technology, Federal Armed Forces University, 85577 Neubiberg/Munich, Germany