Robot vehicle path planning including a tracking of the closest moving obstacle

The problem of the path generation for the autonomous robot vehicle in environment with stationary and moving obstacles is considered. An algorithm, named MKBC, based on modified Kohonen rule and behavioral cloning is developed. The MKBC algorithm, as improvement of RBF neural network, uses the training values as weighting values, rather then values from the previous time. This enables an intelligent system to learn from the examples (operator's demonstrations) to control the robot vehicle, in this case, to track the closest moving obstacle like the human operator does. Important characteristic of the MKBC algorithm is polynomial complexity, while most other path planning algorithm are exponential. Experiments determined that it is robust to parameter change and suitable for real time application.