Autonomous Bayesian escorting of a human integrating intention and obstacle avoidance

Abstract

This paper presents a Bayesian technique for navigating a mobile robot to safely escort a human to their destination. A mathematical analysis of escorting identifies the significance of human intention inference for escorting success. The proposed Bayesian technique observes the head pose of the human walking behind the robot to infer their intention. The future human pose is then predicted by fusing two independent predictions using Gaussian fusion. One prediction method uses the observed intention, and the other prediction uses a physics‐based model. Both prediction methods initially use particle filters to best model nonlinear and non‐Gaussian motion. The robot goal pose is determined from the predicted human motion and adjusted if necessary leveraging the unoccupied distance map to avoid obstacles and successfully and safely enable autonomous escorting. Experimental analysis shows that the incorporation of the proposed human intention model reduces human position prediction error by approximately 33% when turning, improving escorting accuracy by 50%. The incorporation of the proposed obstacle avoidance has further demonstrated its need and efficacy for successful and safe autonomous escorting.

Document Details

Document Type
Pub Defense Publication
Publication Date
Mar 14, 2022
Source ID
10.1002/rob.22070

Entities

People

  • Dean Conte
  • Tomonari Furukawa

Organizations

  • MITRE Corporation
  • Office of Naval Research
  • University of Virginia
  • Virginia Tech

Tags

Fields of Study

  • Computer science

Readers

  • Computational Modeling and Simulation
  • Computer Vision.
  • Exercise and Sports Science.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Bayesian Inference
  • Autonomy