PRISM is a research framework for automated experimental protocol generation, validation, and execution in robotic laboratories. It integrates language-model–based reasoning, simulation-driven validation, and robot-aware execution to enable end-to-end automation without human intervention between experimental steps.
This repository accompanies the PRISM paper and provides the prompts and code used for protocol planning, protocol generation, simulation-based validation, and robotic execution.
Figure: Overview of the PRISM framework for protocol generation and execution.
The system consists of three main stages: Protocol Planning, where user intent is converted into structured steps; Protocol Generation, where structured English instructions are transformed into robot-aware actions and iteratively refined through validation cycles in Omniverse before execution; and Real-World Execution, where the full pipeline is validated using the Luna qPCR protocol in our autonomous laboratory.
PRISM operates as a closed-loop system with three core stages:
User intent is converted into structured natural-language experimental steps using language-model–based reasoning.
This stage may involve:
- Automatically retrieving reference procedures from web-based sources
- Generating structured experimental steps (e.g., liquid handling, timing, dependencies)
- Identifying required reagents, instruments, and constraints
Structured protocol descriptions are transformed into robot-aware, executable protocols.
This stage includes:
- Translation into the Argonne MADSci protocol format
- Coordination across multiple robotic instruments
- Simulation-based validation in a digital twin environment built in NVIDIA Omniverse
- Iterative refinement cycles where detected physical or sequencing errors are reported back and corrected
Protocols must pass simulation-based validation before execution.
Validated protocols are executed on an autonomous laboratory platform composed of off-the-shelf robotic instruments, including:
- Opentrons OT-2 liquid handler
- PF400 robotic arm
- Azenta plate sealer and peeler
The full pipeline is demonstrated using Luna qPCR amplification.
PRISM supports systematic benchmarking across:
- Single-agent vs multi-agent protocol generation
- Constrained vs open-ended prompting paradigms
- Protocol correctness, ordering, and refinement efficiency
Simulation-based validation enables consistent detection of physical infeasibility prior to real-world execution.
This repository is intended for research and benchmarking purposes.
Protocols generated by PRISM should be independently reviewed and validated before use in safety-critical or production laboratory environments.
If you use PRISM or build upon this work, please cite:
@article{prism2025,
title={PRISM: Protocol Refinement through Intelligent Simulation Modeling},
author={...},
journal={...},
year={2025}
}