Sunday, 22 November 2015 09:49

Evolving a walking behaviour for Bioloid robots

Written by

In the course Humanoid Robotics, my group used genetic algorithms to let a Bioloid robot evolve a walking behaviour. The training phase (evolution) took place in the simulation environment V-REP. The robot is controlled by a CPG (central pattern generator) which was optimized using a genetic algorithm.

The following three videos show the evolution of the robot gait in the simulation environment, one of our results in both the simulator and on the real robot, and finally what some of the local optima we found looked like.

Bioloid evolving a gait:

Bioloid walk in reality and simulator:

Bioloid silly walks (evolved using a genetic algorithm):





Bioloid evolving a gait Bjorn Persson
Login to post comments