We present a two-stage approach to tackle early convergence in morphology-controller co-evolution. The new method outperforms traditional co-evolutionary runs by over 10%, and was verified in hardware experiments.
In EvoApplications (Springer), 2017.

We propose using a high level controller with multi-objective optimization on a physical four legged robot to evolve fast and stable gaits. This achieves a range of robust gaits for the quadruped robot, but does not require the excessive hardware tests associated with traditional multi-objective evolution in hardware.
In SSCI (IEEE), 2016.

News mentions, 29. jan 2017
An interview about rapid prototyping and evolutionary robotics (Norwegian), 20. apr 2017
A short talk about the future of robotics (Norwegian), 12. feb 2017 An interview with Minoru Asada and Jim Tørresen about humanoid robots and use of robots in healthcare (Norwegian), 11. oct 2016
Interview about robots that learn, and the goals of the EPEC project (Norwegian), 12. nov 2014
Article about adaptive robots and evolving robot morphologies (Norwegian)



“Engineering Predictability with Embodied Cognition” - A University of Oslo project, funded by the Research Council of Norway

Master thesis

Evolutionary optimization of robot morphology and control - Using evolutionary algorithms in the design of a six legged robotic platform