FM-MRS @ IROS 2026
Foundation Models in Multi-Robot Systems
Overview
Foundation models, including large language models (LLMs) and vision-language models (VLMs), are rapidly transforming robotics by enabling semantic reasoning, language-guided planning, and richer perception capabilities. This workshop focuses on foundation models in multi-robot systems, exploring how large-scale pretrained models can support coordination, decision making, and mission specification across robot teams operating in complex environments.
The workshop highlights emerging directions such as LLM- and VLM-driven multi-robot coordination, heterogeneous robot collaboration, language-specified missions, decentralized reasoning architectures, and reliability mechanisms including constraint checking and conformal prediction. Compared with recent workshops that primarily focus on LLM-enabled robotics, this workshop emphasizes team-level autonomy and system-level scalability, examining how foundation models can facilitate coordination, communication, and planning across multiple robots.
The workshop will bring together researchers from multi-robot systems, robot learning, planning, and large-scale AI models, as well as industry practitioners developing and deploying robotic fleets. Through invited talks, lightning presentations, and interactive discussions, the workshop aims to identify key challenges, share emerging methodologies, and outline future research directions at the intersection of foundation models and multi-robot autonomy.
Invited Speakers
Program Schedule
| Time | Session | Details |
|---|---|---|
| 8:30 – 8:40 | Opening Remarks | Workshop overview and introduction |
| 8:40 – 9:05 | Invited Talk 1 | Chuchu Fan (MIT) — LLMs and VLMs Can Solve Real-World Planning Rigorously with Formal Reasoning Tools |
| 9:05 – 9:30 | Invited Talk 2 | M. Ani Hsieh (University of Pennsylvania) — Physics Foundation Models for Multi-Robot Autonomy: From Team-Level Sensing to Guaranteed-Safe Motion |
| 9:30 – 9:55 | Invited Talk 3 | Javier Alonso-Mora (TU Delft) — Foundation Models for Planner Adaptation and Open-World Perception |
| 9:55 – 10:15 | Contributed Spotlight Talks | Lightning talks from the ten accepted papers |
| 10:15 – 11:15 | Coffee Break & Poster Session | Poster presentations and networking |
| 11:15 – 11:35 | Invited Talk 4 | Jiachen Li (Georgia Tech) — Toward Safe and Efficient Coordination for Cooperative Embodied Agents |
| 11:35 – 11:55 | Invited Talk 5 | Ameya Agaskar (Amazon Robotics) — DEEPFLEET: Multi-Agent Foundation Models for Mobile Robots |
| 11:55 – 12:20 | Panel Discussion | Open Q&A and discussion on future directions |
| 12:20 – 12:25 | Community Announcement | Giuseppe Loianno (UC Berkeley) and Francesco Blasi (ASPIRE) — A2RL Multi-Agent Drone Racing Challenge |
| 12:25 – 12:30 | Closing Remarks |
Accepted Papers
Ten papers were accepted to the workshop. Each is presented as a 2-minute lightning talk during the Contributed Spotlight session, followed by a poster during the coffee break. Papers are available on OpenReview.
- Risk-Bounded Language-to-Task Allocation: A Conformal Prediction Framework for Heterogeneous Robot Teams
- Physical Agentic AI: An Architecture for Orchestrating a Robot Crew with LLMs
- Evaluating Social Reasoning in Embodied Vision-Language Models
- LLM-Guided Adaptive Auction Coordination for Multi-Autonomous Trucks Under Non-Nominal Logistics Yard Conditions
- Multi-Robot Formation Coordination with Verified Agentic LLM Planning
- PIP-LLM: Integrating PDDL-Integer Programming with LLMs for Coordinating Multi-Robot Teams Using Natural Language
- NeuroMesh: A Unified Neural Inference Framework for Decentralized Multi-Robot Collaboration
- TriPlane-WAM: A Multi-Robot World Action Model with Shared Tri-Plane Workspace
- Interactive Grounding for Multi-Robot Coordination using Affordance-Based Hypergraphs
- Verifier-Mediated Multi-LLM Coordination for Multi-Robot Path Planning
Call for Papers
We invite submissions of extended abstracts (2–4 pages, including references) on topics including but not limited to:
- LLM- and VLM-driven multi-robot coordination and planning
- Language-specified missions for heterogeneous robot teams
- Foundation-model-based decentralized reasoning and multi-robot communication
- Reliability and safety of foundation models in robot teams: constraint checking, formal verification, and conformal prediction
- Foundation models for multi-agent decision making and task allocation
- Real-world deployment of foundation-model-enabled robot fleets
Accepted papers are presented as a 2-minute lightning talk followed by a poster presentation during the coffee break session.
Submit your paper via OpenReview.
Please prepare your submission in the standard IEEE conference format. You can find the right template using the IEEE template selector.
Submissions are now closed. The deadline was August 23, 2026.
Organizers
Contact
For inquiries, please contact Lifeng Zhou (lz457#drexel.edu) and Peihan Li (pl525#drexel.edu) at Drexel University.