In Search of a Lost Metric: Human Empowerment as a Pillar of Socially Conscious Navigation

TL;DR

This work addresses the gap in evaluating social navigation by introducing human empowerment, an information-theoretic metric that quantifies how human actions influence their future states in shared spaces. The authors define and estimate empowerment via a variational MI bound using neural networks (Source Policy, Transition, Planning) and ego-centric occupancy maps, enabling tractable computation in dynamic environments. Through CrowdNav-based simulations of multiple policies, empowerment tracks human autonomy and discriminates among policies, with statistical tests confirming significant differences in empowerment across methods. The study demonstrates that empowerment complements existing metrics like discomfort, offering a continuous, interpretable measure of social compliance and human agency with potential implications for policy design and evaluation in human-robot interaction.

Abstract

In social robot navigation, traditional metrics like proxemics and behavior naturalness emphasize human comfort and adherence to social norms but often fail to capture an agent's autonomy and adaptability in dynamic environments. This paper introduces human empowerment, an information-theoretic concept that measures a human's ability to influence their future states and observe those changes, as a complementary metric for evaluating social compliance. This metric reveals how robot navigation policies can indirectly impact human empowerment. We present a framework that integrates human empowerment into the evaluation of social performance in navigation tasks. Through numerical simulations, we demonstrate that human empowerment as a metric not only aligns with intuitive social behavior, but also shows statistically significant differences across various robot navigation policies. These results provide a deeper understanding of how different policies affect social compliance, highlighting the potential of human empowerment as a complementary metric for future research in social navigation.

Figures (8)

• Figure 1: A robot navigating in a crowded environment.
• Figure 2: Network modules of Source, Transition, and Planning used to compute Human Empowerment
• Figure 3: Empowerment for Different Policies vs. Time
• Figure 4: Time-evolution of the robot's trajectory and corresponding empowerment values at different time steps. The robot (blue circle) navigates towards the goal (red star) while interacting with nearby humans (numbered circles). Empowerment, shown in the top of each subplot, decreases as the robot progresses through the environment, reflecting reduced autonomy due to increased proximity to humans. This figure demonstrates how empowerment evolves dynamically in response to the robot's movement through a crowded environment.
• Figure 5: Avg. Empowerment for Different Policies vs. Crowd Sizes