The central pattern generator (CPG) has been found to be a real, existing neuron controller for the locomotion control of animals and it has been used on bio-inspired robots widely in recent years. However, research on the adaptability of CPG-based locomotion control methods is still a challenge. In particular, the performance of the CPG method on quadruped robots is not good enough in some situations compared with the traditional force control methods. In this article, we adopt a CPG method in which phase difference between oscillators can be arbitrarily adjusted, and we try to improve the CPG's applications in quadruped robots in some aspects. One aspect is static walk gait locomotion, in which we try to add a transition state in the CPG network to enhance the static balance of the robot. Another aspect is gait transition. Compared with the traditional abrupt gait transition, we try to realize a continuous gait transition between walk gait and trot gait to decrease the fluctuations of the robot. The improved CPG method is tested on a quadruped model and it shows positive results with regard to the improvement of static walk gait and gait transitions.
- Bio-inspired robots
- Quadruped locomotion
ASJC Scopus subject areas
- Computer Science Applications
- Artificial Intelligence