High-slope terrain locomotion for torque-controlled quadruped robots

Abstract

Research into legged robotics is primarily motivated by the prospects of building machines that are able to navigate in challenging and complex environments that are predominantly non-flat. In this context, control of contact forces is fundamental to ensure stable contacts and equilibrium of the robot. In this paper we propose a planning/control framework for quasi-static walking of quadrupedal robots, implemented for a demanding application in which regulation of ground reaction forces is crucial. Experimental results demonstrate that our 75-kg quadruped robot is able to walk inside two high-slope (50 • ) V-shaped walls; an achievement that to the authors' best knowledge has never been presented before. The robot distributes its weight among the stance legs so as to optimize user-defined criteria. We compute joint torques that result in no foot slippage, fulfillment of the unilateral constraints of the contact forces and minimization of the actuators effort. The presented study is an experimental validation of the effectiveness and robustness of QP-based force distributions methods for quasi-static locomotion on challenging terrain.

Online payload identification for quadruped robots

“…The walking framework is a statically stable crawl framework that we use for rough terrain locomotion. The core module is a state machine (see [13] for details) that switches between two temporized/event-driven locomotion phases: a swing phase, and a body motion phase. During the body motion phase the robot CoM is shifted onto the future support triangle, which is opposite to the next swing leg, in accordance to a user-defined foot sequence.…”

“…This paper is structured as follows: in Section II we start with the generic theory of identification, specified for quadrupeds; Section III presents a state-of-the-art trunk CoM identification method [13] (which is used as ground truth) and two recursive strategies for online identification. Sections IV and V present the data collected from simulation and experiments both for our offline and online identification approaches.…”

“…First we identify the trunk mass. Then Section III-A introduces a state-of-the-art batch Least-Squares approach [13] to identify trunk CoM (approach (a)), which will be used as baseline. In Sections III-B and III-C we describe two recursive strategies for identification of the trunk CoM: one based on the trunk orientation error (approach (b)) and another based on the contact forces (approach (c)).…”

“…This approach is built on top of a trunk posture controller [13] that controls the position of the robot CoM and the orientation of the trunk by means of virtual springs. The controller also compensates for gravity, by applying a force equal and in opposite direction of gravity and passing through the CoM (see Fig.…”

“…Furthermore, arms can be mounted on the trunk for manipulation tasks, creating a centaur [12]. Hence, for trunk inertial parameters, an online identification strategy would be beneficial to identify the changes in the trunk parameters (CoM, mass and inertia) and promptly reflect them into the trunk controller that exploits the rigid body model of the robot to compute the joint torques [13].…”