Data and Code for Post-Disaster Collaborative Road Restoration with Heterogeneous Repair Teams

These inputs represent controlled computational benchmarks, not observations collected from actual disasters. The dataset contains 381 unique optimization-run records: 102 main-comparison runs, 117 component-ablation runs, and 162 additional sensitivity runs. The sensitivity analysis comprises 189 records because it also reuses 27 ablation baselines.

Included materials cover solver-reported objectives, lower bounds, optimality gaps, runtimes, repair modes, team routes, schedules, and validation results, alongside the model implementations, instance generator, summary files, and reproducibility tools. The main comparison uses a 1,800-second optimization limit; ablation and sensitivity runs use a 900-second limit. The main comparison shows that the enhanced formulation reduces the group mean incumbent makespan by 12.55%, from 601.84 to 526.33 for the 20–35-site group, improves incumbents on 23 of 24 instances, and reduces the mean optimality gap from 65.65% to 30.19%.

It also strengthens average lower bounds for the 8–15-site group. Sensitivity-analysis bounds certify that adding a second specialized team strictly reduces the optimal makespan on all three tested 15-site instances. Results should be interpreted as finite-time computational outcomes under synthetic input assumptions.

Incumbent makespans are not necessarily optimal, and missing incumbents do not establish infeasibility. Cumulative synchronization waiting measures team-time rather than an additive increase in restoration makespan. The results do not establish field productivity rates or investment priorities.

The README and data dictionary explain file organization, units, identifiers, and analysis conventions. Included solver-independent tools verify checksums, input changes, saved schedules, and summary statistics without running optimization. Rerunning the optimization requires an appropriately licensed Gurobi installation.

Historical instance codes S and M correspond to the manuscript’s 8–15-site and 20–35-site groups, respectively.