UCLA

This research uncover the internal structure and control mechanisms of oscillatory biological systems, and generalizes synthesis guidelines to embed multiple desired trajectories for robotic design, with analytically determined sensory feedback gain. First we proposed an integrated CPG-leech model amenable for theoretical study, also capable of reproducing adaptive behavior of actual leech undulation in both water and air. Using this model, the internal architecture of CPG was explored, which has never been studied before. The conservative oscillator and weak coupling structure were discovered. In current state of knowledge, the mechanism that how CPGs achieve and maintain orbital stability under perturbations, and how CPGs adjust trajectories under environmental perturbations are unknown to us. This conservative oscillator and weak coupling architecture can well-explain the CPG control mechanism of stabilization and trajectory re-planning. For applications to the community, today's design of CPG controller in robotic system relies heavily on manual tuning of the the sensory feedback or mere open-loop control, lack of established theoretical support. The synthesis also needs to take the complex dynamics of the plant into consideration in the process of determining the CPG matrices, which makes the design computationally inefficient. Based on the newly-found architecture in this research, several design guidelines are generalized to embed multiple targeted orbits, with analytically determined sensory feedback gains. The weak coupling structure allow us to design CPGs only using target oscillation profile, no plant dynamics are required in calculating the connectivity matrix, making the design process direct and efficient.