Enhancing Legged Robot Navigation of Rough Terrain via Tail Tapping

Abstract

Legged systems offer benefits over their wheeled counterparts in their ability to negotiate rugose and heterogeneous environments. However, in rough terrain, these systems are susceptible to failure conditions. These failures can be avoided through different strategies such as environmental sensing or passive mechanical elements. Such strategies come at an increased control and mechanical design complexity for the system, often without the capability of failure recovery. Here, we sought to systematically investigate how contact generated from a tail can be used to mitigate failures. To do so, we developed a quadrupedal C-leg robophysical model (length and width = 27 cm, limb radius = 8 cm) capable of walking over rough terrain with an actuated tail (length = 17 cm). We programmed the tail to perform three distinct strategies: static pose, periodic tapping, and load-triggered (power) tapping, while varying the angle of the tail relative to the body. We challenged the robot to traverse a nearly impassable terrain (length = 160 cm, width = 80 cm) of randomized blocks with dimensions scaled to the robot (length and width = 10 cm, height = 0 to 12 cm). Without the tail, the robot was often trapped among blocks, limiting terrain traversal, independent of gait pattern. With the tail, the robot could still be trapped due to complex interactions (difficult to detect or predict) between the system and its immediate surroundings. However, using the tail, the robot could free itself from trapping with a probability of 0 to 0.5, with the load-driven behaviors having comparable performance to low frequency periodic tapping across all tail tapping angles. By increasing the probability of freeing, the robot was more likely to traverse the rough terrain (mean distance before failure of 1.47 to 2.49 body lengths). In summary, we present a framework that leverages mechanics via tail-ground interactions to mitigate failure and improve legged system performance in heterogeneous environments.