IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026
Recent advances in robot control, perception, planning, and learning have significantly expanded the capabilities of individual robots to operate in unstructured environments and interact with the physical world. As robots move beyond controlled laboratory settings toward real-world deployment, many emerging applications, such as collaborative construction, cooperative manufacturing, and embodied logistics, naturally require multiple robots to coordinate through physical interactions. However, despite growing practical demand, research to enable tightly coupled physical collaboration in multi-robot systems remains underexplored.
Most existing multi-robot research has focused on loosely coupled coordination problems, where robots can be planned largely independently. In contrast, physically collaborative tasks require robots to reason jointly about contact interactions, shared physical constraints, and long-horizon task dependencies, which introduce significant challenges in modeling and computation. Developing coordination methods for physical collaboration will therefore require the community to revisit many of the foundational assumptions underlying multi-robot systems.
This workshop aims to shift the focus of the multi-robot systems community from scalable but loosely coupled coordination toward physically grounded, tightly coupled collaboration. By bringing together researchers from control, motion planning, task and motion planning, and robot learning, the workshop will highlight emerging methods that bridge model-based and learning-based coordination, address long-horizon task dependencies, and integrate single-robot intelligence into multi-robot systems. Through invited talks, peer-reviewed contributions, and panel discussions, the workshop seeks to help define the next research frontier for real-world deployment of multi-robot collaboration.
Robots must reason about complex physical interactions with the environment and with one another, yet overly simplified models fail in practice while high-fidelity models are too expensive for planning and control.
Real-world tasks require reasoning about temporal dependencies, shared resources, and coupled motion constraints. Existing approaches often suffer from the curse of dimensionality or simplify task structure, limiting their ability to handle rich, interdependent objectives.
Coordinated demonstrations are costly to obtain, making training data scarce for multi-robot teams. Meanwhile, teams must handle far greater task diversity, requiring coordination, role assignment, and adaptation to changing environments.
Delft University of Technology
Tentatively remote. Title to be announced.
Peking University
Neural-accelerated task and motion planning for multi-robot systems.
Rice University
Efficient motion planning for manifold-constrained multi-robot manipulators.
Carnegie Mellon University
Scalable planning and asynchronous execution for long-horizon multi-robot manipulation.
University of Michigan, Ann Arbor
Towards site preparation with multi-robot teams.
University of Modena and Reggio Emilia
Tentative: human-aware distributed control of multi-robot systems.
National University of Singapore
High-dimensional multi-agent robot learning.
Researchers from diverse backgrounds, career stages, and research communities will present complementary perspectives aligned with the workshop theme, each followed by a dedicated Q&A.
Researchers from diverse backgrounds and career stages will discuss open challenges to encourage diffusion of ideas and spark new insights.
Each accepted poster will be briefly introduced by its presenter, followed by interactive discussions to engage attendees.
We invite 2-4 page extended abstracts (including references) targeting state-of-the-art research on tightly coupled physical collaboration in multi-robot systems. check more details in the Call for Papers.
Researchers and practitioners in multi-robot systems, motion planning, task and motion planning, robot learning, whole-body control, and embodied AI.