Research Analyzer
← Back ICRA 2026

Teamformer: Scalable Heterogeneous Multi-Robot Team Formation

Noah Boehme, Geoffrey Hollinger

PDF

AI summary

Key figure (auto-extracted from paper)
Teamformer uses a decentralized transformer policy to dynamically form heterogeneous robot teams, scaling linearly to handle up to 100 distinct robot and task types in both simulation and hardware.
Multi-robot task allocation Heterogeneous robots Transformer policy Scalable coordination Decentralized control Team formation

Problem

Existing multi-robot task allocation methods fail to scale when the number of distinct robot capabilities and task requirements increases, creating a computational bottleneck for complex heterogeneous systems.

Approach

The authors formulate a Team Formation Markov Decision Process and train a decentralized transformer policy that uses autoregressive decoding to reduce the team formation action space from exponential to linear growth.

Key results

  • Linear scaling of the action space with distinct robot capabilities via autoregressive decoding
  • Generalization to 100 distinct robot and task types in simulation
  • Decentralized coordination of 20 heterogeneous robots in real-world hardware experiments
  • Competitive performance against optimal centralized solvers in small-scale scenarios

Why it matters

Provides a scalable, decentralized solution for coordinating large, diverse robot swarms in real-world applications like logistics and environmental monitoring.

Abstract

Accounting for heterogeneity among robots and tasks adds additional complexity to multi-robot task allocation. While existing task allocation methods effectively handle het- erogeneity among robots and tasks, they do not scale well in the number of different robots and tasks. To address this gap, we formulate the Team Formation Markov Decision Process (TF-MDP) for training Teamformer: a scalable, decentralized transformer policy for dynamically forming heterogeneous teams of robots to complete diverse tasks. Combining the TF- MDP with the autoregressive capability of transformers enables Teamformer to scale linearly in the number of robots, tasks, and combinations of different heterogeneous robots. Simulations demonstrate Teamformer generalizing to combinations of 100 different types of robots and tasks. Hardware experiments using Georgia Tech’s Robotarium show Teamformer decentrally coordinating up to 20 heterogeneous robots for task completion.

Index terms

Multi-Robot Systems Reinforcement Learning

Related papers