AAAI 2027 Workshop
Coordinating AI Scientists for Autonomous Physical Experimentation
About the workshop
Over the past two years, autonomous “AI Scientist” agents have shown they can search literature, write code, run computational experiments and draft papers with little human help. That work has now moved from digital to physical environments: robotic agents act in real laboratories and run autonomous experiments in chemistry, biology, materials science and beyond.
A real autonomous laboratory is never one agent. It is a system of heterogeneous agents coordinated by a shared world model of the laboratory. A domain-expert agent proposes hypotheses, a planner agent turns them into executable protocols, and embodied robot agents carry those protocols out. The world model tracks their interactions, predicts what each step should produce, detects failures when observations deviate, and feeds corrections back to the responsible agent.
Self-driving labs such as Berkeley’s A-Lab already work this way. Its AlabOS orchestration layer synthesized around 3,500 samples over 1.5 years. But that coordination is hand-engineered and must be rebuilt for every new lab. This workshop looks at the physical frontier of auto-research. We will survey laboratory automation across fields, compare hand-engineered orchestration with emerging agentic and world-model-driven approaches, and identify what is still missing on the path to generally autonomous laboratories.
Call for papers
We invite contributions on a broad range of topics, including but not limited to:
Role decomposition and communication patterns for physical experimentation.
Task allocation and multi-agent planning on shared instruments under temporal, resource and safety constraints.
VLA/VLM-based manipulation, navigation and perception; grounding natural-language protocols into verified robot-executable actions.
Monitoring and review agents that check results against action logs, sensor readings and instrument data, and catch mismatches between reported and actual execution.
Oversight interfaces, intervention protocols, and dividing scientific authority between human researchers and agent collectives.
Simulators, digital twins and remote/cloud-lab platforms for evaluating and comparing multi-agent lab systems across labs.
Deployments in chemistry, biology, materials science and beyond, analyzed through coordination, prediction and verification, not domain results alone.
Safety constraints for physically acting agents, authorship and accountability, and governance of increasingly autonomous labs.
Both tracks are hosted on OpenReview. We welcome work that is unpublished or currently under submission elsewhere.
Full papers of up to seven pages, excluding references and supplementary material, which have no page limit.
Works in progress and intermediate milestones. We especially encourage submissions from underrepresented, under-resourced and early-career researchers to share experiences, get feedback and find collaborators.
Every paper gets at least three experienced reviewers. Conflicts of interest are handled through OpenReview.
Reviewers can nominate papers from either track for an oral. Non-conflicted reviewers then rank the nominees.
An award for a paper authored or co-authored by a student, to recognize their research potential and raise their visibility.
Important dates
The AAAI-27 workshop program takes place Monday, February 22 and Tuesday, February 23, 2027, at the Palais des Congrès de Montréal. Deadlines follow the AAAI-27 workshop program guidelines. All deadlines are 23:59 Anywhere on Earth (AoE).
Program
A full-day, in-person program with invited talks, four oral presentations, two poster sessions and a panel discussion.
Confirmed
Expertise spanning self-driving laboratories, autonomous chemistry and materials experimentation, synthetic biology and biofoundries, and operations management.
University of Toronto, Canada
Professor of Chemistry and Computer Science and Director of the Acceleration Consortium. Combines AI, robotics and high-throughput experimentation to build materials-acceleration platforms and self-driving laboratories.
University of Glasgow, UK
Regius Chair of Chemistry, leading research in digital chemistry and programmable synthesis. His group develops the Chemputer and the chemputation framework, which translate and execute chemical procedures and integrate AI with robotic experimentation.
University of Toronto, Canada
Professor of Materials Science and Engineering and Research Scientist at CanmetMATERIALS. Leads the Autonomous Discovery of Alloys (AutoDIAL) group, building AI-guided automated platforms for materials discovery, with interests in autonomous lab science and trustworthy AI.
Aalto University, Finland
Professor in the Department of Industrial Engineering and Management. His research spans operations and supply-chain management, intelligent products, tracking and tracing, and technology-enabled innovation.
Air Force Research Laboratory, USA
Autonomous Materials Lead in AFRL’s Materials & Manufacturing Directorate. Developed ARES, one of the earliest autonomous research systems for materials, which combines robotics, AI and data science to direct and run experiments autonomously.
Wright State University, USA
Assistant Professor of Computer Science and Engineering and Director of the Knowledge and Semantic Technologies (KASTLE) Lab. Works on knowledge graphs, ontology modeling and design patterns, and neurosymbolic AI.
University of Tennessee, Knoxville, USA
MathWorks Professor at the University of Tennessee, Knoxville. Works on high-performance and scientific computing, numerical reproducibility, large-scale scientific data analytics, and computing infrastructure for AI-driven scientific discovery.
National University of Singapore
Provost’s Chair Professor of Biochemistry and Executive Director of the National Centre for Engineering Biology, Singapore. Brings extensive biofoundry experience in automated biological experimentation and large-scale research infrastructure.
University of Illinois Urbana-Champaign, USA
Steven L. Miller Chair Professor of Chemical and Biomolecular Engineering. Integrates synthetic biology, machine learning and lab automation. His group built iBioFAB, which couples computational design with robotic execution of biological workflows.
Organizing committee
An intentionally cross-disciplinary committee from universities, national laboratories and industry, spanning agentic AI, robotics, chemistry, materials science and synthetic biology.
Main organizers
EPFL, Switzerland
Professor in EPFL’s Laboratory of Molecular Simulation. Works on molecular simulation, computational chemistry, and data-driven modeling for molecular and materials discovery.
Imperial College London, UK
Professor at Imperial College London, Director of the London BioFoundry and Co-Director of SynbiCITE. Works on synthetic biology, biofoundries, genome engineering, and automated biological experimentation.
Aalto University, Finland
Assistant Professor of Operations Management. Works on agentic AI scientists, autonomous research, vision–language models for cyber-physical systems, and AI-enabled production and operations management.
Oak Ridge National Laboratory, USA
Staff Scientist at ORNL. Works on AI for science, scientific hypothesis generation, large language models for science, agentic AI for discovery, and AI for laboratory operations.
Tulane University, USA
PhD candidate working on reinforcement learning, foundation models for robotics, embodied AI, and autonomous research.
AI, robotics & autonomous scientific systems
Georgia Institute of Technology, USA
Stephen Fleming Early Career Professor in Computer Science. Works on robot learning, embodied AI, reinforcement learning, foundation models for robotics, and self-driving laboratories.
NVIDIA, USA
Senior Business Development Manager. Focuses on AI for science, high-performance scientific computing, AI-accelerated simulation, and integrating simulation with experimental data.
Chalmers University of Technology, Sweden
Professor of Data Science and AI, known for the Robot Scientists Adam and Eve. Works on automated scientific discovery, laboratory robotics, and closed-loop autonomous experimentation.
Chemistry, materials science & molecular discovery
NIST, USA
Materials Research Engineer. Works on applied AI, scientific machine learning, active learning, materials characterization, and autonomous experiment planning and execution.
University of Wuppertal, Germany
Professor of Digital Chemistry. Works on machine learning for chemical reactivity, Bayesian optimization, reaction prediction, and data-driven reaction optimization.
HEC Montréal, Canada
Associate Professor working on AI for science, geometric deep learning, generative models for molecule and protein discovery, and generative agents for biomedicine.
Synthetic biology & biological automation
Genome Foundry, Concordia University, Canada
Senior Advisor for Business Development and Partnerships. Works on genome foundries, microbial genome engineering, and automated high-throughput biological experimentation.
Stanford University, USA
Professor of Bioengineering and, by courtesy, Chemical Engineering. Works on synthetic biology, cell-free bioengineering, AI-guided biotechnology, and biomanufacturing.
Contact
For questions about submissions or the program, contact: