Learned Fast Rolling for Tensegrity Robot

These are initial results of using evolutionary learning algorithms to develop a rolling gait for a tensegrity robot. The robot has an open-loop distributed control system where each string has a sine-wave controlling its length, and the learning algorithm is tuning the basic parameters such as phase off-sets, amplitude, and frequency. The tensegrity robotics simulator used is based on the Bullet Physics Engine and was developed at NASA Ames Research Center. More information can be found at