Event

Authors propose auditable self-driving laboratory framework

1 reports · 1 sources

Zihan Zhang and other authors propose on arXiv a traceable self-driving laboratory framework consisting of four components—LAM, OpenSDL, Safe-SDL, and L3-Bench—designed to connect experiment planning, automated execution, and measurement feedback. Among them, LAM is a measurement-conditioned policy architecture that extends the vision-language-action paradigm and simultaneously generates continuous robot trajectories and typed instrument API calls; OpenSDL serves as the execution substrate, has a preliminary control implementation, and is responsible for mapping role-level operations to heterogeneous devices, coordinating execution, and recording timestamped causal events; Safe-SDL is a runtime gating mechanism that combines pre-dispatch schema and resource checks with signal temporal logic monitoring during execution (including extended waits). L3-Bench specifies a set of episode-level evaluation protocols covering physical outcomes, timing, reproducibility, safety violations, and failure recovery. All the above components are designs proposed by the authors, and the report does not provide actual deployment or validation results.