Machine Learning Methods for Multi Robot Navigation

In multi robot navigation, robots need to move towards their goal positions while adapting their paths to account for potential collisions with other robots and static obstacles. Existing methods compute motions that are optimal locally but do not account for the motions of the other robots, producing inefficient global motions when many robots move in a crowded space. In my research approach, each robot uses online machine learning techniques to adapt dynamically its behavior to the local conditions. The approach is highly scalable because each robot makes its own decisions on how to move. Experimental results obtained in simulation under different conditions show that the robots reach their destinations faster using motions that are more energy efficient.