sdl-workflows: hardware workflows of closed-loop self-driving laboratories

Hardware workflows of the closed-loop self-driving laboratories (SDLs) described in 237 published papers. For each of the 279 platforms, the data list its modules, the steps one experimental cycle runs on them, and the connections that move material or data between modules. Every item has a class from a fixed taxonomy and the page of the source paper that describes it. The times, batch sizes, throughputs and costs the papers state are attached to the coded items: 454 step durations, 185 numbers of samples one device handles in one run of a step, and 472 values of platforms, modules and connections, such as batch sizes, operating hours, measured or nominal throughput and transfer times. For each paper, the data also record how it reports human intervention during its closed-loop campaign. Every table is given as CSV and Parquet files and in one SQLite database, with a data dictionary, a data card that describes how the papers were found and coded, the taxonomy, the coding guideline, the extraction instructions, and a benchmark for predicting module types and transfer modes of papers from 2024 on from papers up to 2023, with a scoring script. The data contain no text of the papers. Explorer: https://ccy5123.github.io/sdl-workflows-explorerCode: https://github.com/ccy5123/sdl-workflows