Data and code for Lag-Aware Tension Allocation for Cable-Driven Parallel Robot Tracking With Differentiable Reinforcement Learning

Recorded data and source code accompanying Lag-Aware Tension Allocation for Cable-Driven Parallel Robot Tracking With Differentiable Reinforcement Learning, by Jing Zhu (Hohai University; ORCID 0009-0007-1297-3677). The archive includes 2,376 primary simulation episodes, 4,644 dynamic-allocation comparison episodes, 11,340 CPU allocator-call timing measurements, training histories for five seeds, 25 policy state dictionaries, boundary/derivative diagnostics, Python training and analysis scripts, and editable Blender/Draw.io figure assets. The README provides installation and reproduction commands. The data dictionary lists CSV fields and units; a manifest and SHA-256 hashes identify the files. The included verification script checks case indices, seed coverage, key manuscript aggregates and tension-limit records. Report scripts were run against the archived CSVs, and a 600-step representative trajectory replay was checked against its recorded commands and position errors. Version 1.0.0. Data, checkpoints, documentation and original figure assets: CC BY 4.0. Python source code: MIT, as specified in LICENSE-CODE. CPU timing measurements describe the recorded host; environment JSON files document the hardware and numerical settings.