Detecting Surface Features During Locomotion using Optic Flow

We test the hypothesis that: (1) Optic flow can be used to detect significant environmental features during locomotion in a biped, even given significant up and down movement and jarring of the robot during locomotion. (2) Reliable detection is only possible if a prediction of the expected optic flow field is made at each instance. This prediction should be driven by the phase of the robot's gait as well as other information about the state of the robot. (3) This prediction can be accomplished in a distributed, biologically plausible framework. Our results using a walking biped mechanism strongly support this hypothesis and indicate that optic flow is a viable strategy and that the prediction of optic flow is a critical component in this behavior. We believe this is the first time that this approach has been used and demonstrated successfully on a walking machine. For walking robots to realize their full potential, distal environment sensing must be tightly integrated with the walking cycle. Distal sensing is crucial to allow anticipatory gait adjustment to accommodate varying terrain. Close coupling of the visual and locomotor cycle can lead to rapid, adaptive adjustment of the robot. Biological systems have a tightly integrated action perception cycle, unlike that seen in most legged robots. In the 1950's, Gibson pointed out the importance of the flow of the 'optic array' (i.e. optic flow), a visual cue arising from relative movement of the environment, in the control of human locomotion [1]. Relative motion of the environment can reveal environment structure. His pioneering ideas have influenced generations of experimentalists in the brain and behavioral sciences. In robotics, the route to using optic flow has come principally from studies of insects. Insect motion processing pathways are relatively well understood. There have been a number of robotic implementations that use optic flow as its primary input for visual navigation. For a flavor of that work, see [2-9]. Vision is a particularly elegant, multifunctional distal sense. It is not surprising that biological systems have relied so heavily on vision to guide locomotion. However, there has been little or no work using optic flow to control legged locomotion although there has been interest in other visual techniques for the control of walking machines (visual servoing, color, stereopsis, terrain reconstruction using laser ranging). Given the strong belief in the biological community that optic flow is a key component of visuomotor coordination (e.g. [10-12]), …