PACC: A Passive-Arm Approach for High-Payload Collaborative Carrying with Quadruped Robots Using Model Predictive Control

TL;DR

The paper addresses the payload bottleneck in legged robots by introducing a passive arm with intrinsic impedance (PACC) that augments carrying capability without actuators. It couples a simple 3-DoF passive arm design with a decentralized Model Predictive Controller that incorporates arm dynamics and estimated end-effector forces to coordinate leader-follower carrying on rough terrain. Key contributions include the mechanical design of a lightweight passive arm, an MPC formulation that accounts for payload-arm coupling, and experimental validation across robot-robot and human-robot collaboration scenarios on stairs and irregular terrain. This approach enables high-payload collaborative carrying with improved safety, robustness, and terrain adaptability, while reducing actuator complexity and energy expenditure.

Abstract

In this paper, we introduce the concept of using passive arm structures with intrinsic impedance for robot-robot and human-robot collaborative carrying with quadruped robots. The concept is meant for a leader-follower task and takes a minimalist approach that focuses on exploiting the robots' payload capabilities and reducing energy consumption, without compromising the robot locomotion capabilities. We introduce a preliminary arm mechanical design and describe how to use its joint displacements to guide the robot's motion. To control the robot's locomotion, we propose a decentralized Model Predictive Controller that incorporates an approximation of the arm dynamics and the estimation of the external forces from the collaborative carrying. We validate the overall system experimentally by performing both robot-robot and human-robot collaborative carrying on a stair-like obstacle and on rough terrain.

Figures (9)

• Figure 1: Experimental scenarios to assess three different collaborative carrying tasks using PACC: on the top-figure, robot-robot CC with rigid payload coupling; on the mid-figure, robot-robot CC with non-rigid payload; and on the bottom-figure, human-robot CC with rigid payload coupling.
• Figure 2: PACC mechanical design and components: a) the arm has 3 revolute joints, arranged in a yaw-pitch-pitch configuration; b) the location of the springs and kinematics of the arm (see Table I for the data referring to the green dots). Note that joint 1 and 3 can also be equipped with an antagonistic pair of springs instead of only one spring (as seen in the figure); c) each joint is equipped with an encoder to measure angular displacements; d) structural elements, highlighted in yellow, create friction forces against the spring extension and retraction to insert damping. The 4 subfigures show the arm in the joint zero-position configuration. Kinematic data related to the green dots are described in Table \ref{['tab:kin_params']}.
• Figure 3: Velocity command zones according to the arm joint displacements: on the left, lateral view showing the angular ranges used to obtain the desired robot forward velocity, where $\theta$ is the orientation of Joint 3 with respect to the gravity vector; on the right, top view showing the angular ranges used to obtain the desired robot's heading velocity, where $\psi$ is the angular position of Joint 1.
• Figure 4: Sequence of numbered snapshots from experimental tests of collaborative carrying on a stair-like obstacle and rough terrain. The two quadruped robots are endowed with the passive arm to carry a 7kg payload connected to a stiff bar in a leader-follower manner. From snapshots 1 to 6, the robots walk up and down the stairs and turn around to cross the rocks. The stair obstacle is composed of pallets that are 55cm in depth and 16cm (bottom pallets) and 13cm (top pallet) in height.
• Figure 5: Plots for the experimental tests with two robots carrying a 7kg payload attached to a rigid bar. The left column refers to the period in which the robots walk on the stairs while the right column when traversing the rocks. From top to bottom, the first plot shows the forward velocity commands for each robot. The following two plots show the estimated forces along the longitudinal axis and moments on the sagittal plane of the trunk horizontal frame. The last two plots show the ZMP margin for each robot and the dotted line is the corresponding margin constraint.