Flexible and multistable pattern generation by evolving constrained plastic neurocontrollers

摘要: In evolutionary robotics, plastic neural network models proved to be promising for evolving adaptive behaviors. particular, neurocontrollers incorporating hebbian synapses have been shown useful implementing conflicting sub-behaviors. Numerous interesting complex tasks assume such flexibility. However, those evolved controllers often exhibit behavioral instability, as simulation time is extended beyond the short limit used during evolution. this paper, we propose constrained inspired by homeostasis phenomena, in order evolve flexible and stable pattern generators single-legged locomotion. Comparative results show that perform better than unconstrained ones both terms of evolvability stability. Functional analyses best controller unveil adaptivity, robustness arising from statically plasticity. Interestingly, implicitly without relying on any active homeostatic mechanisms implemented through plasticity, usually considered destabilizing.