EvoRBC

EvoRBC implementation for OpenAI Gym environments

References

1. Illuminating search spaces by mapping elites (MAP-Elites)
2. Evolution of Repertoire-Based Control for Robots With Complex Locomotor Systems (EVO-RBC)
3. Evolving Neural Networks through Augmenting Topologies (NEAT)
4. Muscle contributions to fore-aft and vertical body mass center accelerations over a range of running speeds

Ant observations/todos after 3rd repo

1. some legs still don't move in a few gaits, maybe increase control freq range but then can't stay still.... (maybe penalise this.. difference in leg speed)
2. the topmost speed isn't fast enough could increase the range
3. evolved behaviors are good.
4. since things have saturated and mutation isn't leading anywhere great, should start including crossover

EvoRBC high level architechture

Control function for joints used for hexapod

In the ant implementation, frequency is also taken as variable parameter since it helps in varying velocity

MAP Elites algorithm

Unified QD algorithm

Regarding cross over in map elites according to the paper

Activation and excitation for muscles - human

Muscles activations during one gait cycle