Sensor-driven hexapod robot with terrain classification for adaptive gait control and 2D time-of-flight mapping

This chapter presents the development of a sensor-driven hexapod robot equipped with a perception and control framework for terrain-adaptive locomotion. A custom hexapod platform was designed and built, incorporating 18-DOF leg actuation, onboard electronics, and a full gait-execution system capable of running multiple bio-inspired locomotion patterns. To support terrain perception, a compact 2D time-of-flight (ToF) sensor and an onboard vision pipeline were developed. A dedicated dataset of terrain textures was collected and used to train a Convolutional Neural Network (CNN) for real-time terrain classification. The vision system processes the lower camera view using homography projection and patch-based analysis to identify terrain categories directly from the robot’s perspective. Based on the predicted terrain type, the control framework automatically selects the most suitable gait from three options—Tripod, Wave, or Bipod—to optimize stability and movement efficiency. Experimental evaluations validate the complete system, demonstrating reliable perception, accurate terrain classification, and effective terrain-aware locomotion on the developed hexapod robot.