A compact reformulation of the two-stage robust resource-constrained project scheduling problem: Complete results

Published: 12-05-2020| Version 1 | DOI: 10.17632/86mxp9c9bf.1
Matthew Bold


Complete results corresponding to the research presented in the paper entitled 'A compact reformulation of the two-stage robust resource-constrained project scheduling problem', submitted to 'Computers and Operations Research' in April 2020 (Bold and Goerigk, 2020). The uncertain resource-constrained project scheduling problem (RCPSP) test instances on which these results are obtained are derived from deterministic instances in the PSPLIB (http://www.om-db.wi.tum.de/psplib/) involving 30 activities (j30). 3 robust counterparts corresponding to the delaying of (Gamma=) 3, 5 and 7 activities respectively have been generated for each of the 480 deterministic PSPLIB instances. Hence, a total of 1440 test instances have been generated. This data set contains four data files corresponding to the results of four variants of the model proposed in Bold and Goerigk (2020). These are a 'basic' model (basic_results_full.txt), and three extended models: including transitivity constraints (trans_results_full.txt), warm-start (warmstart_results_full.txt), transitivity constraints plus warm-start (warmstarttrans_results_full.txt). See Bold and Goerigk (2020) for details of these methods. Each data set reports for each instance the instance number (corresponding to the deterministic PSPLIB instance), the number of activities delayed (Gamma), the Gurobi solution status code, the best lower and upper-bound found by the solver, the gap between these two values, and the CPU run-time for that instance. Results show that the proposed model out-performs the current state-of-the-art algorithms for solving the two-stage robust resource-constrained project scheduling problem, being much quicker to solve, and reaching optimality for 50% more instances on the same benchmark set.