Elastic string based global path planning using neural networks

We present an efficient potential-field based path planning approach for point or higher dimensional objects which avoids effectively any local extreme problems. The approach represents a path by a series of via points connected by elastic strings which are subject to displacement due to collisions with obstacle regions as well as constraints pertaining to the domain to which path planning is applied. Obstacle regions are represented by a potential field created by a multilayered neural network. A fast simulated annealing approach is used for local minima problems from the potential field. The automatic generation and removal of via points is incorporated in the path planning approach to ensure collision-free planning regardless of the complexity of the environment. Local and global bias on the potential field is used to avoid any existing singularities and local minima. Our path planning approach is flexible, efficient and massively parallel