Obstacle avoidance for multi-legged robot using virtual impedance model

An obstacle avoidance method using a virtual impedance wall is proposed for a multi-legged robot. The swing legs use compliance control to make soft contact and avoid colliding with objects, so that the robot maintains the moving direction as far as possible while the swing legs maintain a preferred operating region. The preferred operating region is surrounded with a virtual impedance wall. When the leg passes over the threshold of the preferred operating region, the moving direction is modified by the virtual repulsive force from the virtual impedance wall to avoid the workspace limitation. Moreover, the pattern recognition technique using the support vector machine is implemented for estimating the contact point between the object and the body by using the data set of the error of each leg. The virtual impedance field is set at the estimated contact position to direct the aspect of avoidance. The robot is pushed and rotated by the virtual repulsive force from the impedance field. These passive motions from the virtual impedance model can provide a good solution for object avoidance control. The feasibility of the proposed obstacle avoidance method is shown by experiments using actual robots. © 2013 The Japan Society of Mechanical Engineers.