Continuous assessment of gait stability in limit cycle walkers

Walking in the presence of perturbations poses a demanding challenge to bipedal robots. In this context, we propose an approach based on an algorithm that observes continuously the state of the robot, and if a perturbation is detected, triggers a recovery action. This paper focuses in the former, it presents a simple algorithm to detect online the occurrence of a perturbation with short delay. Simulations with a complex biped model yield no false positives when walking stably, and provide detection of trips with an average delay of 30.2 ms, comparable to short latency reflexes in humans.