Final Year Project

Project Title

Quadruped Locomotion in Rough Terrain and Confined Environments

Description

Legged robots, in contrast to wheeled robots, have the added benefit of legged locomotion, allowing them to tackle different terrains (caves, tunnels, and rubble) that a wheeled robot cannot. Using a hobby-grade quadruped robot, we augmented it with mounts for our compute modules and sensors to obtain feedback required for robust locomotion. For locomotion in rough terrain, we tested classical control, learning-based approaches, and evolutionary algorithms. To allow the robot to navigate through confined spaces, we used a depth camera for perceiving the immediate terrain and deciding on a posture to constrain the locomotion algorithms to minimize collisions. Thanks to our simulation package, we were able to rapidly test and evaluate our algorithms virtually and physically using a testbed.