Problem
Research questions and friction points this paper is trying to address.
embodied navigation
human routines
long-term reasoning
human-centered environments
behavioral patterns
HUMEMBR: Learning Human Routines for Predictive Embodied Navigation
🤖 AI Summary
This work addresses the challenge of enabling robots to understand and predict human behavior over extended periods of cohabitation for effective localization and navigation. The authors propose a novel embodied agent system that, for the first time, integrates long-term human behavioral memory with embodied navigation. By continuously constructing structured routine representations, employing a parallel retrieval mechanism, and supporting user-driven interactive queries, the system achieves efficient, low-overhead long-horizon behavior reasoning. Evaluated on a real robotic platform across two environments, the approach accurately responds to diverse queries and successfully completes conditional navigation tasks. It significantly outperforms full-context large language model baselines while drastically reducing the number of tokens required.
📝 Abstract
Understanding and navigating human-centered environments over extended periods of time while considering human behavior and routines remains a fundamental challenge in robotics. In real-world settings, robots may be asked to locate a specific individual, predict where that person is likely to be, or estimate when they typically leave a building. Addressing such queries requires reasoning over extensive histories of observations and capturing long-term behavioral patterns. To this end, we introduce Human-Centered Memory for Embodied Robots (HUMEMBR), a system designed for embodied question answering and routine-conditioned navigation. HUMEMBR integrates a continuous memory construction process with a parallel retrieval and querying mechanism, enabling the system to accumulate structured representations of human routines while supporting interactive, user-driven queries. Our experimental results indicate that HUMEMBR improves long-horizon reasoning about human behavior relative to full-context LLM baselines, while using substantially fewer tokens. Furthermore, we deploy HUMEMBR on a physical robot in two distinct environments, showing its ability to handle diverse queries and navigation tasks under real-world conditions.
Innovation
Methods, ideas, or system contributions that make the work stand out.
embodied navigation
human routines
memory system
long-horizon reasoning
interactive querying
🔎 Similar Papers
No similar papers found.
