Research on fusion control of sensor information and biological reflection based on CPG

Abstract

To enhance the environmental adaptability of the quadruped robot and realize its stability in terrain with obstacles, this paper studies the fusion of sensor information feedback and bio-reflection control algorithm. First, a central pattern generator (CPG) that generates basal rhythmic movements is constructed. Secondly, a simplified foot end that can be used to identify the touchdown and obstacle information is designed. To enhance the stability of the quadruped robot during the flexor reflex process and avoid the problem of solidification in the reflex process, the foot touch signal was introduced into the flexor reflex algorithm, which enhanced the flexibility of the reflex process. To avoid secondary collisions with obstacles, the postswing reflection process is optimized, and active obstacle avoidance is realized after the obstacle is touched. Finally, the organic fusion of sensor information feedback, CPG control network, and bio-reflection control algorithm is realized. The simulation results show that the robot’s ability to overcome obstacles is improved, the number of reflections is less, and the reflection process is more stable.

Central pattern generators evolved for real-time adaptation to rhythmic stimuli

“…Early bio-inspired robotics work attempted to replicate this emergent pattern generation, typically with continuous-time recurrent neural networks [16,17]. This connectionist approach was largely abandoned in favour of more manageable coupled limit-cycle oscillators, where each parameter's effect on the overall behaviour is predictable [1,18,19], particularly if the CPG outputs are mapped to workspace trajectories [20]. Hence, the problem of gait adaptation has become a matter of designing appropriate feedbacks or learning schemes for the designated control parameters.…”