Perception and state estimation play an important role to achieve breakthroughs in dynamic whole-body locomotion for quadruped robots.
This means that legged robots should be able to perceive their surrounding environment, detect any sudden change, and take decisions based on that.
The detection aspect in quadruped robots includes (but is not limited to): detecting and localizing contact points between the robot and an obstacle, detecting the terrain's impedance properties, and detecting slippage.
Based on that, this thesis focuses on developing a new approach for slip detection, the first fundamental step to then implement a robust controller that allows the robot the locomotion on slippery ground.
Once the robot detect slippage, it can trigger a reflex action and adapt it's motion trajectories to not slip.
We validated the proposed approach on the 90 Kg Hydraulically actuated Quadruped robot HyQ (Istituto Italiano di Tecnologia), and we compared it against a state-of-art slip detection algorithm.
We showed that our approach can result in a better slip detection, which is gait and robot's base states (that are drifting) independent.