Complementary videos for the thesis: "A Modular Robot Architecture for Learning to Move and Be Automatically Reconfigured", in review.
Controller with sensors evolved using the locomotion training framework.
Controller with sensors evolved using the locomotion training framework with Hill Climbing.
Controller with no sensors (CPG coordination mechanism only) evolved using the locomotion training framework with HAEA.
Controller with sensors evolved using the locomotion training framework with HAEA for a T-Shaped Morphology.
Controller with sensors evolved using the locomotion training framework with HAEA. The initial population comes from previously obtained controllers with no sensors (HAEA-NS-CPG).
Controller with sensors evolved using the locomotion training framework with HAEA. The initial population comes from a manually designed controller without sensors.
Controller with sensors evolved using the locomotion training framework with HAEA and a short challenge incremental approach.
Controller with sensors evolved using the locomotion training framework with HAEA and a short challenge incremental approach, for a T-Shaped morphology.
Controller with sensors generated with the locomotion training framework in simulation and transferred to reality, tested in the straight primitive environment.
Controller with sensors generated with the locomotion training framework in simulation and transferred to reality, tested in the right corner primitive environment.
Testing the locomotion training framework in reality: Examples of the best controllers from before evolving, after evolving and after rearranging the environment (Generalization).