Robot Path Finding Using Neural Net
Moaid A. Fadhil · 2005
To act in an unknown and stationary environment, robot must be able to react instantaneously to changes and unexpected events in order to avoid collisions with obstacles, and to update its map. The objective of this research is to design, implement, and evaluate a new model of robot path finding in a stationary environment by using one of reinforcement learning algorithms called Q-learning algorithm. The proposed model has three goals; first, is to run the proposed model efficiently over robot stationary environment, by exploring the entire unknown environment. Second, is to train robots coordinate a behavior according to instant situations in a stationary environment. Third, is to estimate an optimal path which is safe and short path, from starting point to goal point. The achieved results and aims of Q-learning algorithm can be classified as follows: 1. The important achieved aim of Q-learning algorithm at stationary environment is: Coordinate the behaviors of the robot by get the right decision in the stationary environment. 2. The achieved aim of Q-learning algorithm at speed level is to find a number of paths in a second, which has proved system model performance. 3. The achieved aims of Q-learning algorithm at exploration efficiency level are: a) Robot explores the entire unknown environment in which other search methods like heuristic search