Exploiting Natural Dynamics in the Control of a Planar Bipedal Walking Robot

Natural dynamics can be exploited in the control of bipedal walking robots: the swing leg can swing freely once started; a kneecap can be used to prevent the leg from inverting; and a compliant ankle can be used to naturally transfer the center of pressure along the foot and help in toe off. Each of these mechanisms helps make control easier to achieve and results in motion that is smooth and natural looking. We describe a simple control algorithm using these natural mechanisms which requires very little computation. The necessary sensing consists of joint angles and velocities, body pitch and angular velocity, and ground reaction forces. Using this simple algorithm, we have controlled a seven link planar bipedal robot, called Spring Flamingo, to walk. Video, photographs, and more information on Spring Flamingo can be found at http://www.leglab.ai.mit.edu