Cheap Design Approach to Adaptive Behavior: Walking and Sensing through Body Dynamics

Abstract — The approach of cheap design has been providing a number of implications by viewing behavior as the result of a system-environment interaction. It states that adaptive behavior cannot be reduced into the problem of control architectures, but the coupling between the system and the environment is essential. Case studies in the past have demonstrated not only the efficient behaviors but also the underlying mechanisms in the perception and the learning processes. This paper explores additional insights of the cheap design approach by introducing two recent case studies of the locomotion robots, which involve a few potentially important research directions for our further understanding of adaptive behavior. The first case study investigates the complex leg operation during bipedal walking. A newly developed biped robot will demonstrate the relatively complex human locomotion as a selforganized behavior of the compliant passive legs without the necessity of sensory feedback. In the second case study, we will discuss how the body dynamics influences the cognitive processes; It will be shown that, by exploiting the intrinsic body dynamics, the sensor information can be meaningfully structured for the purpose of situated perception of the environment. I