Gait Synthesis in Legged Robot Locomotion Using a CPG-Based Model

Biology has always been a source of inspiration and ideas for the robotics community. Legged locomotion problem is not an exception, and many experiences have taken ideas from animals, both for morphological and behavioral issues. The first ideas for gait generation came from animal observation, but they were mainly focused on mimicking legs movements. It was not until the nineties that the first relevant works appeared trying to identify the principles behind the generation of those movements in animals. The proposed models were based on neurophysiologic principles, and most of them tried to include characteristics of animal locomotion by the addition of neural networks, dynamic oscillators, or using a set of “movement rules”. Although many models have been suggested, most of them share some common aspects: 1. Motion signals generation and processing are very slow and highly distributed processes. 2. The brain tends to perform high level feed-forward movement control and prediction. 3. The locomotion system has local feedback, from pressure sensors, force sensors, intramuscular sensors, etc. In some processes these characteristics are obvious, like in the heart beating or breathing. In these processes there is no need for the intervention of a complex processing unit like the brain, since most of the coordinated oscillatory behavior of the muscles is carried out locally and distributed. The oscillatory nature of locomotion patterns has attracted studies about the existence of a similar structure in charge of this problem. The biologic and electrochemical bases of the system in animals are fairly well explained in the works on neural networks by Hodgkin-Huxley (Hodgkin, 1952). Another important characteristic of animal systems is that biological neural networks can perform timing tasks through oscillatory networks, and also can modulate neuromuscular excitatory signals, thus giving the ability to