N-Learning: an approach to autonomous learning and teching in multi-robot systems
Multi-Robot System, Autonomous Robots, Behavior-Based Robotics, Learning by Social Interactions.
We propose the N-Learning, an architecture that provides learning in autonomous Multi Robot Systems. Our approach allows the robots to learn new behaviors in time execution. The N-Learning is Behavior-Based Robotics and based on Learning by Social Interactions. The behaviors in a robot are represented by a graph. Complex behaviors can be decompose into simpler behaviors that run simultaneously in the robot. An advantage of N-Learning is that is possible to change the robot domain without reprogram the robots. A robot that have behaviors for a specific domain can change it and learn with the other robots. This advantage is useful when exists a large number of robots. We developed an implementation of N-Learning based on Python language and Robotics Operating System. Experiments were performed in Stage simulator and with real robots in order to validate our approach.