When Do LLM Agents Help? Deadline-Aware Mixed-Criticality Task Scheduling at the Autonomous-Vehicle Edge
3.80T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2608.19557.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryStudy comparing a heuristic windowed contract-net auction scheduler with a multi-agent LLM control layer for deadline-aware mixed-criticality task scheduling on mobile edge computing servers for autonomous vehicles. Across 60 instances, the heuristic reached 0.902 time-critical completion rate. The LLM layer only added significant value during a mid-run surge of safety-critical tasks, not under stationary load.
Why it mattersA controlled decomposition isolating when LLM orchestration actually earns its overhead, with code and instances released. Calibrates expectations for deploying LLM agents in scheduling roles.
Cited by
No citations on record.