AgentComm-Bench: Stress-Testing Cooperative Embodied AI Under Latency, Packet Loss, and Bandwidth Collapse

arXiv cs.AI / 2026/3/24

📰 ニュースSignals & Early TrendsIdeas & Deep AnalysisTools & Practical UsageModels & Research

共有:

要点

The paper introduces AgentComm-Bench, a benchmark suite and evaluation protocol that stress-tests cooperative embodied AI under realistic network impairments such as latency, packet loss, bandwidth collapse, asynchronous updates, stale memory, and conflicting sensor evidence.
It evaluates cooperative embodied tasks across three families—cooperative perception, multi-agent waypoint navigation, and cooperative zone search—using five communication strategies including a proposed redundant message coding approach with staleness-aware fusion.
Experimental results show catastrophic degradation for communication-dependent behaviors, with navigation performance dropping by over 96% under stale memory and bandwidth collapse and perception F1 falling by over 85% under corrupted or conflicting data.
The study finds impairment vulnerability is task- and mechanism-dependent, e.g., perception fusion is robust to packet loss but can amplify corrupted inputs, while redundant message coding more than doubles navigation performance under 80% packet loss.
The authors release AgentComm-Bench publicly and recommend that cooperative embodied AI research report results under multiple impairment conditions rather than only idealized communication assumptions.

Abstract

Cooperative multi-agent methods for embodied AI are almost universally evaluated under idealized communication: zero latency, no packet loss, and unlimited bandwidth. Real-world deployment on robots with wireless links, autonomous vehicles on congested networks, or drone swarms in contested spectrum offers no such guarantees. We introduce AgentComm-Bench, a benchmark suite and evaluation protocol that systematically stress-tests cooperative embodied AI under six communication impairment dimensions: latency, packet loss, bandwidth collapse, asynchronous updates, stale memory, and conflicting sensor evidence. AgentComm-Bench spans three task families: cooperative perception, multi-agent waypoint navigation, and cooperative zone search, and evaluates five communication strategies, including a lightweight method we propose based on redundant message coding with staleness-aware fusion. Our experiments reveal that communication-dependent tasks degrade catastrophically: stale memory and bandwidth collapse cause over 96% performance drops in navigation, while content corruption (stale or conflicting data) reduces perception F1 by over 85%. Vulnerability depends on the interaction between impairment type and task design; perception fusion is robust to packet loss but amplifies corrupted data. Redundant message coding more than doubles navigation performance under 80% packet loss. We release AgentComm-Bench as a practical evaluation protocol and recommend that cooperative embodied AI work report performance under multiple impairment conditions.