Toward simple control for complex, autonomous robotic applications: combining discrete and rhythmic motor primitives

摘要: Vertebrates are able to quickly adapt new environments in a very robust, seemingly effortless way. To explain both this adaptivity and robustness, promising perspective neurosciences is the modular approach movement generation: Movements results from combinations of finite set stable motor primitives organized at spinal level. In article we apply concept generation movements control robots with high number degrees freedom, an issue that challenging notably because planning complex, multidimensional trajectories time-varying laborious costly process. We thus propose decrease complexity phase through use combination discrete rhythmic primitives, leading decoupling (i.e. choice behavior) actual trajectory generation. Such implementation eases of, switch between, different behaviors by reducing dimensionality high-level commands. Moreover, since generated dynamical systems, can be smoothly modulated, either commands change current behavior or sensory feedback information environmental constraints. order show generality our approach, framework interactive drumming infant crawling humanoid robot. These experiments illustrate simplicity architecture terms planning, integration types (vision contact) capacity autonomously switching between (crawling simple reaching).