Closing the loop: Evolving a model-free visually guided robot arm

Dynamic neuro-controllers are incrementally evolved for reaching and tracking movements by a physically simulated robot arm. An active vision system capable of controlling gaze direction and focus replaces the need for internal models of the robot. It is shown that closing the feedback loop allows for robot control being robust to changes in environment, sensors and robot-morphology.

[1]  Zbigniew Michalewicz,et al.  Evolutionary Computation 1 , 2018 .

[2]  Zoubin Ghahramani,et al.  Computational principles of movement neuroscience , 2000, Nature Neuroscience.

[3]  Lina L. E. Massone.  Sensorimotor learning , 1998 .

[4]  Dario Floreano,et al.  An evolutionary active-vision system , 2001, Proceedings of the 2001 Congress on Evolutionary Computation (IEEE Cat. No.01TH8546).

[5]  Zoubin Ghahramani,et al.  Computational motor control , 2004 .