Modular Neural Control for Gait Adaptation and Obstacle Avoidance of a Tailless Gecko Robot

Abstract

In this study, we present a novel neural control architecture for gait adaptation and obstacle avoidance of a tailless gecko robot. The control architecture is based on a hierarchical modular structure, consisting of several neural layers and modules. The first layer contains three sensory preprocessing modules which filter sensory noise and generate appropriate descending commands to activate corresponding behaviors through the second and third layers. The second and third layers contain a central pattern generator (CPG) module and CPG postprocessing modules, respectively. The CPG module generates basic rhythmic locomotion patterns, shaped by the CPG postprocessing modules to achieve different gaits (i.e., wave, intermediate, and trot) as well as different climbing directions (i.e., forward and sideways). We use a body inclination sensor to adapt the robot gait while climbing on different slope angles, with infrared sensors to detect an obstacle on its climbing path and activate obstacle avoidance behavior. We successfully tested our control approach on a real tailless gecko robot. As a result, the robot can efficiently climb forward on different slope angles (including 90∘) and automatically adapt its climbing gait accordingly, to maximize climbing speed and ensure stability. It can also avoid an obstacle by changing its climbing direction from forward to sideways.