Risk-Calibrated Human-Robot Interaction via Set-Valued Intent Prediction

RSS 2024

Abstract

Tasks where robots must cooperate with humans, such as navigating around a cluttered home or sorting everyday items, are challenging because they exhibit a wide range of valid actions that lead to similar outcomes. Moreover, zero-shot cooperation between human-robot partners is an especially challenging problem because it requires the robot to infer and adapt on the fly to a latent human intent, which could vary significantly from human to human. Recently, deep learned motion prediction models have shown promising results in predicting human intent but are prone to being confidently incorrect. In this work, we present Risk-Calibrated Interactive Planning (RCIP), which is a framework for measuring and calibrating risk associated with uncertain action selection in human-robot cooperation, with the fundamental idea that the robot should ask for human clarification when the risk associated with the uncertainty in the human's intent cannot be controlled. RCIP builds on the theory of set-valued risk calibration to provide a finite-sample statistical guarantee on the cumulative loss incurred by the robot while minimizing the cost of human clarification in complex multi-step settings. Our main insight is to frame the risk control problem as a sequence-level multi-hypothesis testing problem, allowing efficient calibration using a low-dimensional parameter that controls a pre-trained risk-aware policy. Experiments across a variety of simulated and real-world environments demonstrate RCIP's ability to predict and adapt to a diverse set of dynamic human intents.

Goal

In the setting where a robot has access to an intent prediction model (e.g., a VLM) and human help, we aim to achieve both: (1) risk calibration: the robot should seek sufficient help to ensure a statistically guaranteed level of task success specified by the user while also satisfying a limit on the human help rate, and (2) flexible autonomy: the robot should be able to tune its behavior (e.g., fewer errors with more help) by providing a range of prediction parameters that control its level of risk. We collectively refer to these sufficiency and minimality conditions as certifiable autonomy.

Approach

RCIP builds upon statistical risk calibration (SRC) to formally quantify and bound multiple notions of risk in human-robot interaction (HRI). Using a small set of calibration scenarios, RCIP computes step-wise prediction losses to form an aggregate emperical risk estimate. Using a risk limit, for each pair (λ,θ) of prediction thresholds and tunable model parameters, RCIP evaluates the hypothesis that the test set risk is above the limit. Thus, for all hypotheses that are rejected, the test set risk satisfies the threshold (with high probability).

RCIP can be applied in multi-step settings using novel extensions to statistical risk calibration derived in our work. RCIP provides a flexible framework for optimizing the episode-level task completion and human intervention rates on both sequence and step levels. KnowNo, which generates plans in open-ended language, may generate a plan that is technically correct, but ambiguous to execute for a language-conditioned policy (both the blue and white bin have a pot). RCIP instead guarantees that the human’s intent is satisfied via constraint satisfaction with the intent-conditioned planner

RCIP also imbues prediction models with task knowledge through a discrete set of human intents. An open-ended VLM planner may generate a plan that is technically correct, but ambiguous to execute for a language-conditioned policy. RCIP instead guarantees that the human’s intent is satisfied via constraint satisfaction with the intent-conditioned planner.

Citation


                    @article{lidard2024risk,
                      title={Risk-Calibrated Human-Robot Interaction via Set-Valued Intent Prediction},
                      author={Lidard, Justin and Pham, Hang and Bachman, Ariel and Boateng, Bryan and Majumdar, Anirudha},
                      journal={arXiv preprint arXiv:2403.15959},
                      year={2024}
                    }