Evolution of behaviors in autonomous robot using artificial neural network and genetic algorithm

摘要

The following introductory basic and explanatory comments are focus setting: 1. Multi-robot systems differ from single-robot systems most significantly in that other robots can affect the environment's dynamics. 2. The pursuit system contains predator robots and preys. 3. Each robot contains sensors to perceive other robots in several directions and decides its behavior based on the information obtained from the sensors. 4. The predator robots catch the preys escaping from them through evolution. 5. The prey has only pre-knowledge to identify and avoid robots in four directions. 6. For the simulator of the system, the structure of the robot (predator/prey) or the behavior decision controller is modeled with the artificial neural network. 7. The genetic algorithm (GA) is used for robot learning, thus giving intelligence to predators in pursuing movement towards the prey. 8. Various selection methods of GA are considered for simulation. 9. A virtual environment is designed with 20 robots and a prey. Simulation was used to: 1. Verify the validity of the system. 2. Observe emergent behaviors. The paper reports several interesting results of the model and the simulation exercises.