抄録
In recent years, CPG (Central Pattern Generator) modelled on animals is attracting attentions as a suitable control technique for legged robots. The CPG network is considered to be effective to make the legged robot capable of adjusting its gait on uneven terrain. Gait-transition is an indispensable subject because it is necessary to choose a suitable gait in uneven terrain. However, the conventional technique only changes the topology of the CPG network, and it does not necessarily guarantee a feasible gait after transition. In this paper we propose a new method of valuing parameters of neuron oscillator, and show the proposed technique is effective for the four legged robot.
