Human–Robot Collaborative Reaction Development with a Self-Driving Laboratory
5 affiliations
- 1Flow Chemistry Group
- 2Van ’t HoInstitute for Molecular Sciences, University of Amsterdam
- 3Chemistry Department, University of Bari Aldo Moro, Via Orabona 4, Bari, Italy, 70125
- 4Group Innovation & Sustainability, Testing, Analytics and Physics Group
- 5Zuyd University of Applied Sciences
Abstract
Self-driving laboratories (SDLs) can accelerate chemical optimisation, yet their application has largely focused on established reaction classes rather than the uncertain and iterative process of developing new reactivity. Here we introduce human-robot collaborative reaction development, in which researchers drive ideation, interpretation and scientific decisionmaking, while an SDL performs the systematic experimentation required to rapidly evaluate emerging reaction concepts. We demonstrate this approach through the development of a photochemical α-(fluoro)alkylation of cyclic secondary amines via in situ imine formation. Starting from an initial human-discovered reactivity hit, RoboChem-Flex was repeatedly deployed as new questions emerged: to establish whether the reaction warranted further development, adapt conditions to di erent substrate classes, rapidly assess coupling partners, and optimise promising but underperforming reactivity. Transfer of knowledge between campaigns further enabled an initially low-yielding α-methylation to be developed in only 11 additional experiments. The resulting methodology accommodates diverse cyclic amines and alkyl iodides and enables downstream derivatisation. More broadly, this work demonstrates how autonomous experimentation can augment human scientific reasoning, accelerating the transition from initial reactivity to synthetic methodology while keeping scientific judgement at the centre of the discovery process.
Related
- A sustainability-aware self-driving laboratory
- Collaborative Laboratory Edge Networks
- Bridging the Literature-to-Laboratory Gap with Language Models and Active Learning for Reaction Feasibility Prediction and Autonomous Exploration
- Combining physical models with dynamically acquired experimental information for the optimization of multicomponent NASICON fast ionic conductors in a self-driving laboratory
- A photochromic reaction to design original laboratory sessions for studying kinetics and uncertainties.
BibTeX
@misc{vanzella2026humanrobot,
title = {{Human–Robot Collaborative Reaction Development with a Self-Driving Laboratory}},
author = {Michael Vanzella and Prakash Chandra Tiwari and Elisabeth Reischauer and Junsong Liu and Marek Pontvuijst and Rossella Catanzaro and Simone Pilon and Elia Savino and Méritxell Fillols and Antonio Pulcinella and Stefano Bonciolini and Ron A.H. Peters and Tim den Hartog and Antonio Monopoli and Timothy Noel},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15008166/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008166/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Human–Robot Collaborative Reaction Development with a Self-Driving Laboratory
AU - Michael Vanzella
AU - Prakash Chandra Tiwari
AU - Elisabeth Reischauer
AU - Junsong Liu
AU - Marek Pontvuijst
AU - Rossella Catanzaro
AU - Simone Pilon
AU - Elia Savino
AU - Méritxell Fillols
AU - Antonio Pulcinella
AU - Stefano Bonciolini
AU - Ron A.H. Peters
AU - Tim den Hartog
AU - Antonio Monopoli
AU - Timothy Noel
PY - 2026
DA - 2026/09/01
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008166/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008166/v1
AB - Self-driving laboratories (SDLs) can accelerate chemical optimisation, yet their application has largely focused on established reaction classes rather than the uncertain and iterative process of developing new reactivity. Here we introduce human-robot collaborative reaction development, in which researchers drive ideation, interpretation and scientific decisionmaking, while an SDL performs the systematic experimentation required to rapidly evaluate emerging reaction concepts. We demonstrate this approach through the development of a photochemical α-(fluoro)alkylation of cyclic secondary amines via in situ imine formation. Starting from an initial human-discovered reactivity hit, RoboChem-Flex was repeatedly deployed as new questions emerged: to establish whether the reaction warranted further development, adapt conditions to di erent substrate classes, rapidly assess coupling partners, and optimise promising but underperforming reactivity. Transfer of knowledge between campaigns further enabled an initially low-yielding α-methylation to be developed in only 11 additional experiments. The resulting methodology accommodates diverse cyclic amines and alkyl iodides and enables downstream derivatisation. More broadly, this work demonstrates how autonomous experimentation can augment human scientific reasoning, accelerating the transition from initial reactivity to synthetic methodology while keeping scientific judgement at the centre of the discovery process.
ER -