Skip to main content

Evolutionary approaches for legged robot control

Posted by: kot01b

February 9, 2017

Due to their morphologies, legged robots are ideal platforms for investigating biologically-inspired approaches to control and navigation. We are currently investigating the application of evolutionary/machine learning techniques to generate task-specific and platform-specific controllers, targeting improved performance. As an example, depending on the control system, variation of joint controller gain values provide a way to decrease energy consumption during operation. We present a fully automated hardware optimisation test-bed that use Evolutionary Algorithms to find a optimal set of controller parameters that increase locomotion performance.

Baldwin on testbed

A testbed that evolves hexapod controllers in hardware

Related papers

  • Huub Heijnen, David Howard, Navinda Kottege, (2017), A Testbed that Evolves Hexapod Controllers in Hardware, To appear in proceedings of the IEEE International Conference on Robots and Automation (ICRA 2017), Singapore, May 2017.
    [Abstract][Paper] Download citation [BibTex]