The N-Body Problem: Parallel Execution from Single-Person Egocentric Video

TL;DR

This paper formalizes the N-Body Problem in egocentric videos, proposing a structured prompting method for Vision-Language Models to generate feasible parallel activity plans, significantly improving action coverage and reducing conflicts.

Contribution

It introduces the N-Body Problem framework, new evaluation metrics, and a structured prompting strategy for VLMs to reason about multi-agent task execution from a single egocentric video.

Findings

Action coverage increased by 45% for N=2

Collision rates reduced by 55%

Object and causal conflicts reduced by 45-55%

Abstract

Humans can intuitively parallelise complex activities, but can a model learn this from observing a single person? Given one egocentric video, we introduce the N-Body Problem: how N individuals, can hypothetically perform the same set of tasks observed in this video. The goal is to maximise speed-up, but naive assignment of video segments to individuals often violates real-world constraints, leading to physically impossible scenarios like two people using the same object or occupying the same space. To address this, we formalise the N-Body Problem and propose a suite of metrics to evaluate both performance (speed-up, task coverage) and feasibility (spatial collisions, object conflicts and causal constraints). We then introduce a structured prompting strategy that guides a Vision-Language Model (VLM) to reason about the 3D environment, object usage, and temporal dependencies to produce a viable parallel execution. On 100 videos from EPIC-Kitchens and HD-EPIC, our method for N = 2 boosts action coverage by 45% over a baseline prompt for Gemini 2.5 Pro, while simultaneously slashing collision rates, object and causal conflicts by 55%, 45% and 55% respectively.