Implicit Communication in Human-Robot Collaborative Transport

TL;DR

This work addresses the challenge of implicit coordination in human-robot collaborative transport by modeling a space of joint workspace traversal strategies and embedding strategy uncertainty into control. The core method, IC-MPC, uses a probabilistic inference of the unfolding traversal strategy from observed joint actions and introduces an entropy-based cost to actively reduce partner uncertainty while maintaining task efficiency. The approach is formalized through a homotopy/topology-based representation of traversal strategies and implemented on a mobile manipulator, achieving higher task success and more fluent robot collaboration than baselines in a lab study (N=24). The findings highlight the practical potential of communicating through object-velocity signals to negotiate safe, efficient collaboration, with implications for broader physically embodied HRI tasks and real-time strategy inference.

Abstract

We focus on human-robot collaborative transport, in which a robot and a user collaboratively move an object to a goal pose. In the absence of explicit communication, this problem is challenging because it demands tight implicit coordination between two heterogeneous agents, who have very different sensing, actuation, and reasoning capabilities. Our key insight is that the two agents can coordinate fluently by encoding subtle, communicative signals into actions that affect the state of the transported object. To this end, we design an inference mechanism that probabilistically maps observations of joint actions executed by the two agents to a set of joint strategies of workspace traversal. Based on this mechanism, we define a cost representing the human's uncertainty over the unfolding traversal strategy and introduce it into a model predictive controller that balances between uncertainty minimization and efficiency maximization. We deploy our framework on a mobile manipulator (Hello Robot Stretch) and evaluate it in a within-subjects lab study (N=24). We show that our framework enables greater team performance and empowers the robot to be perceived as a significantly more fluent and competent partner compared to baselines lacking a communicative mechanism.

Figures (7)

• Figure 1: Footage from our study ($N=24$) involving the collaborative transport of an object (orange stick) by a user and a mobile manipulator in a workspace with an obstacle (red color). The robot runs our controller (IC-MPC), designed to balance functional and communicative actions in collaborative tasks.
• Figure 2: A human (H) and a robot (R) collaboratively move an object from an initial pose $p_0$ to a final pose $g$ in a workspace $\mathcal{W}$. An obstacle $\mathcal{O}$ stands in their way. To avoid collisions with $\mathcal{O}$ and reach $g$, they have to coordinate on a strategy of workspace traversal. In this work, we engineer implicit coordination through the velocities $a$ and $u$ that the human and the robot exert on the object.
• Figure 3: Illustration of our topological abstraction for representing strategies of workspace traversal.
• Figure 4: Users experienced three different starting configurations with each robot algorithm. From left to right: the user and the robot stand side-by-side; the user stands directly behind the robot; the user stands directly in front of the robot. For the third configuration, the user faces toward the goal and holds the object behind their back.
• Figure 5: Workspace traversal strategy inference. The prior distribution for $\mathbb{P}(\textsc{left} \mid p, c)$ is shown as a colormap in the background, and the mode of the action likelihood distribution for $\mathbb{P}(a \mid \textsc{left}, p, c)$ is shown using gray arrows in the foreground.