Mutual Adaptation in Human-Robot Co-Transportation with Human Preference Uncertainty

TL;DR

A unified framework to address the uncertainty of human preference parameters and the need to balance adaptation strategies that benefit both humans and robots is proposed and validated with real human feedback.

Abstract

Mutual adaptation can significantly enhance overall task performance in human-robot co-transportation by integrating both the robot's and human's understanding of the environment. While human modeling helps capture humans' subjective preferences, two challenges persist: (i) the uncertainty of human preference parameters and (ii) the need to balance adaptation strategies that benefit both humans and robots. In this paper, we propose a unified framework to address these challenges and improve task performance through mutual adaptation. First, instead of relying on fixed parameters, we model a probability distribution of human choices by incorporating a range of uncertain human parameters. Next, we introduce a time-varying stubbornness measure and a coordination mode transition model, which allows either the robot to lead the team's trajectory or, if a human's preferred path conflicts with the robot's plan and their stubbornness exceeds a threshold, the robot to transition to following the human. Finally, we introduce a pose optimization strategy to mitigate the uncertain human behaviors when they are leading. To validate the framework, we design and perform experiments with real human feedback. We then demonstrate, through simulations, the effectiveness of our models in enhancing task performance with mutual adaptation and pose optimization.

Figures (6)

• Figure 1: (a) Task: Human-Robot Co-Transportation. (b) Environment: Navigating through obstacles to the target. There exist multiple path options to reach. In some options, there is a high risk of collision (e.g., $A \rightarrow \xi_{1} \rightarrow \xi_{4} \rightarrow B$), or with some options, the team needs to traverse a longer distance (e.g., $A \rightarrow \xi_{3} \rightarrow B$).
• Figure 2: True risk $(\gamma=1)$ and true distance $(\alpha = 1)$. For perceived risk, when $\gamma<1$ the weighting function overemphasizes small risks and underemphasizes large risks, whereas when $\gamma>1$ the opposite trend is observed. Perceived distance varies with $\alpha$, reflecting differences in individual sensitivity.
• Figure 3: Overview of our proposed mutual adaption framework in human-robot co-transportation with human preference uncertainties. The Human Preference Model produces a probability distribution of options. A time-varying human stubbornness measure is developed that incorporates these probabilities into a Human-Robot Coordination Model to determine their coordination modes. If human leads, the Pose Optimization, informed by these models, generates joint configurations that better prepare the robot's low-level control to handle potential human uncertain behaviors.
• Figure 4: With the environment in Fig. \ref{['environment']}, the probability distribution of choosing an option based on the parameter distributions $\alpha\sim\mathrm{U}(0.55, 0.95)$ and $\gamma\sim\mathrm{U}(0.5, 1.1)$ with $c_\text{r} = 10$. The colored regions correspond to the lowest subjective cost of each option. The probabilities are computed based on the areas of these regions.
• Figure 5: Start position $A$, target position $B$. (a) Environment 1: Options: {$(A, \xi_2, \xi_1, \xi_3, B)$, $(A, \xi_2, \xi_4, B)$, $(A, \xi_1, \xi_4, B)$, $(A, \xi_3, B)$}. (b) Environment 2: Options: {$(A, \xi_2, \xi_4, \xi_6, B)$, $(A, \xi_1, \xi_3, \xi_6, B)$, $(A, \xi_2, \xi_7, B)$, $(A, \xi_1, \xi_5, B)$.} (c) Environment 3: Options: {$(A, \xi_4, B)$, $(A, \xi_2, \xi_1, B)$, $(A, \xi_2, \xi_3, B)$, $(A, \xi_4, \xi_3, \xi_1, B)$.} (d) Environment 4: Options: {$(A, \xi_4, B)$, $(A, \xi_3, B)$, $(A, \xi_2, B)$, $(A, \xi_1, B)$}. We use this figure to collect human responses, where the trajectories through openings are not visualized so that participants do not get any extra information.