[Blue Sky] Designing for Long Horizon Social Interactions

Ifrah Idrees

Keywords: Long-Horizon Planning, Long-Horizon State Estimation, Human-Robot Interaction, Assistive Robotics

TL;DR: This is Blue Sky Paper proposing work that aims to bridge the gap in enabling the robot to perform accurate online inference of the world and user state in a partially observable environment, especially over extended periods.

Abstract: Conversational assistive robots have the potential to guide humans to accomplish various daily tasks such as cooking meals, performing exercises, or operating machines. However, to interact effectively over long-horizon tasks, the robot must be able to infer the human's goal from their interactions with the surrounding environment. A few key challenges in inferring the world and user state are that the sensors are noisy, and the robot has partial observability of the environment and the human's intention. These problems compound as the robot collects more noisy observations about the user and environment over long periods of time. Our proposed work aims to bridge the gap in enabling the robot to perform accurate online inference of the world and user state in a partially observable environment, especially over extended periods.