Development of a Hardware CPG Model for Controlling a Musculoskeletal Humanoid Robot to Generate Gait Patterns for Both Legs

Humanoid robots are important research targets because they imitate humans and can be easily imagined as an alternative to the labor force. Conventional humanoid robots are mainly controlled by numerical calculations using CPUs. However, realizing real-time motion control requires a high-performance CPU with high power consumption due to the need for high-speed processing of a huge amount of numerical calculations. This leads to the problem of overloading the on-board battery. In response to this problem, there is the concept of imitating human motion control, since humans are active with energy as low as 100[W] in terms of electricity. Based on this concept, our study has focused on artificial neural networks composed of analog electronic circuits to mimic the central pattern generator (CPG) of the neural network involved in human gait control without using numerical calculations. The objective of this paper is to develop a hardware CPG model for controlling both legs of a musculoskeletal humanoid robot, engineering a mimicry of human muscles and skeleton. Circuit simulations of the constructed hardware CPG model were performed, and it was confirmed that a gait pattern was generated for both legs.