Tese e Dissertação

Tese: Detection and Analysis of Slip in Quadruped Robots

Quadruped robots have many applications in the field of mobile robotics. They operate in rugged environments that challenge the capabilities of wheeled or tracked robots. However, they demand higher computational power and accurate environmental perception to enable their autonomous operation. Slippage is particularly challenging for legged systems, occurring when the ground reaction force (GRF) violates the friction cone. Slippage affects state estimation, self-localization, control systems, and increases the risk of falling. One approach to address this issue is slip detection, which triggers a reaction strategy. However, among the methods available in the literature, there is little direct comparison between them, and there is potential to incorporate additional slip characteristics to assist in reaction and improve environmental understanding. This dissertation aims to compare proprioceptive slip detection methods. The comparison uses simulation data, analyzing results through graphs and quantifying true and false detections. A filter is proposed to correlate a robot's foot's tangential and orthogonal velocity to filter out undesired detections. The study also investigates slip direction, proposing an estimator for its angle. An exponential moving average filter with a single iteration is applied to smooth the measurements. The proposed approach is validated in simulations with the Go1 robot navigating on slip-prone scenarios such as icy surfaces, inclined planes, and flat terrain. The comparison of the slip detection methods highlights the limitations of each approach and the relationship between true and false positives. At the same time, the proposed filter, in simulation, successfully filters out undesired detections. The filtered slip angle shows smoothness, consistent convergence in both simulated and experimental data, and a correlation with slip intensity.

Link da defesa:

https://puc-rio.zoom.us/j/93322186355?pwd=eP01NA3obBW4jrBknc8P6g4rfrwi6V.1

Ver todos