Beyond Fixed Goal Delivery: Online POMDP Planning for Target Interception in Crowds

2026-07-20

AI Summary

This work addresses the challenge of robotic interception of moving targets in dense human crowds, where high uncertainty arises from unobservable human intentions. The problem is formulated as a partially observable Markov decision process (POMDP), and online tree search is employed for planning under a fixed computational budget. The study compares sequential path–velocity planning with a unified joint steering–velocity planning approach, revealing inherent structural limitations of sequential planning in high-density scenarios. To overcome these shortcomings, the authors propose a joint steering–velocity planning method that integrates both control dimensions within a single optimization framework. Experimental results in simulated environments with up to 200 pedestrians demonstrate that the proposed method significantly outperforms baseline approaches, achieving a 31-percentage-point increase in safe interception rate and a 44% reduction in task completion time at the highest crowd densities.

Abstract

Target interception in crowded environments requires reaching a moving objective while navigating among multiple uncertain human agents. Since human navigation intent is not directly observable, the robot must reason over multiple possible future interaction outcomes. We formulate interception in crowds as a partially observable Markov decision process and solve it online using tree search under a fixed computational budget. In this setting, the action-space structure directly shapes the search tree and how computational effort is allocated. We perform a controlled comparison between a sequential path-speed planner, which first plans a spatial path and then modulates speed along it, and a unified planner that jointly branches over steering and speed within tree search. Across simulations with up to 200 humans, both approaches perform similarly at low crowd density but diverge sharply as density increases. At the highest crowd density, the sequential planner has a safe-interception rate 31 percentage points lower and requires 44% more time than the unified steering-speed planner, revealing a structural limitation of spatial restriction. Project webpage: https://tic-planning.github.io/

Problem

• target interception
• crowded environments
• partially observable Markov decision process
• human-robot interaction
• online planning

Innovation

• online POMDP planning
• target interception
• unified steering-speed planning
• crowd navigation
• tree search