Year: 2026
Type: preprint
Abstract: Self-driving laboratories are transforming experimental science, shifting laboratory operation from human-centric schedules towards continuous robotic execution. Beyond accelerating discovery, this transition creates new flexibility in when scientific research consumes electricity, offering an opportunity to coordinate experimental activity with changing electricity-system conditions. Here, we develop and deploy a dynamic, sustainability-aware laboratory-level scheduling framework that combines reinforcement learning with constraint programming to jointly optimize completion time and electricity cost under scientific constraints. Physical execution of 36 experimental tasks spanning six chemical workflow categories shortened completion time by 39.1% and reduced estimated electricity costs by 28.9% relative to a simulated human-centric scheduling baseline. Across five representative tariff cases, estimated per-workflow savings were equivalent to the cost of 0.83–9.14 kWh of off-peak electricity. Our findings demonstrate how self-driving laboratories can translate operational freedom into electricity-demand flexibility, establishing energy responsiveness as a design consideration for scientific research infrastructure.
Source: ChemRxiv
Institution: University of Science and Technology of China
Language: English
Topic: Smart Grid Energy Management
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Awards: G4457991146, G4630919444, G8359409411
OpenAlex ID: W7223288082