Internal State Estimation in Groups via Active Information Gathering

TL;DR

This work tackles the problem of estimating hidden internal states, such as personality, in groups by introducing an active information-gathering framework that couples a personality-conditioned navigation model (via the Eysenck PEN theory) with a receding-horizon planner and Bayesian inference. A learned, personality-conditioned human local planner and a learned navigation simulator enable scalable, differentiable prediction of crowd behavior; a Model Predictive Path Integral (MPPI) controller steers the robot to probe humans while balancing exploration, collision avoidance, and information gain. Across simulations, neurotypical adults, and autism-focused preliminary studies, the approach reduces personality prediction error and uncertainty, demonstrates scalability to tens of humans, and shows potential for ASD-related applications. The results suggest that active information gathering can provide rich, actionable insight for interactive robotics in group settings, paving the way for ASD-specific interventions and broader improvements in human–robot collaboration. Future work will extend to multi-modal interaction, real-robot deployments, and integration of multiple internal states, with a stronger emphasis on safety and comfort in diverse environments.

Abstract

Accurately estimating human internal states, such as personality traits or behavioral patterns, is critical for enhancing the effectiveness of human-robot interaction, particularly in group settings. These insights are key in applications ranging from social navigation to autism diagnosis. However, prior methods are limited by scalability and passive observation, making real-time estimation in complex, multi-human settings difficult. In this work, we propose a practical method for active human personality estimation in groups, with a focus on applications related to Autism Spectrum Disorder (ASD). Our method combines a personality-conditioned behavior model, based on the Eysenck 3-Factor theory, with an active robot information gathering policy that triggers human behaviors through a receding-horizon planner. The robot's belief about human personality is then updated via Bayesian inference. We demonstrate the effectiveness of our approach through simulations, user studies with typical adults, and preliminary experiments involving participants with ASD. Our results show that our method can scale to tens of humans and reduce personality prediction error by 29.2% and uncertainty by 79.9% in simulation. User studies with typical adults confirm the method's ability to generalize across complex personality distributions. Additionally, we explore its application in autism-related scenarios, demonstrating that the method can identify the difference between neurotypical and autistic behavior, highlighting its potential for diagnosing ASD. The results suggest that our framework could serve as a foundation for future ASD-specific interventions.

Figures (13)

• Figure 1: The pipeline of our system (left) and the neural network structure of our human behavior model (right). During each timestep, the robot observes a short history of neighboring human trajectories, from which beliefs over human personalities are updated using Bayesian rules. Finally, our robot planner optimizes the robot control policy to actively probe human personalities.
• Figure 2: Two environments used in experiments. Left: an empty room; Middle: a ring-like environment. Right: the corresponding environment in 3-D simulation. The red disk-shaped agent with circular sensor range in 2-D simulation and the red human in 3-D simulation represent the robot.
• Figure 3: Human behaviors generated by different models with different personalities. In each case, a crowd (brown) with the same control parameters is initially located on the left side and trying to move to the right side.
• Figure 4: Interactive behavior of the robot.
• Figure 5: The number of observed interactions changing history under active and inactive controllers.