- Update (2018-11-05): MIPLIB 2017 is released. The Collection dataset includes 1 Group 1 instance (lr2-22dr3-333vc4v17a-t60), 3 Group 2 instances (lr1dr02vc05v8a-t360, lr1dr04vc05v17a-t360, lr1dr12vc10v70b-t360), and 1 Jiang-Grossmann instance (maritime-jg3d9). The main conclusion is that state-of-the-art commercial mixed-integer programming solvers continue to struggle when solving Group 1 and 2 MIRPLib instances without fit-for-purpose algorithms. The four Group 1 and 2 MIRPLib instances have all been classifed as "open," meaning that no state-of-the-art solver has declared these instances solved to provable optimality. In fact, the objective function values reported are vastly inferior to those reported on the MIRPLib website.
- Update (2018-09-28): Friske and Buriol found 4 new best known solutions to Group 1 instances: LR1_1_DR1_4_VC3_V11a and LR1_1_DR1_4_VC3_V12b for T=45 (i.e., a planning horizon of 45 periods).
- Update (2017-09-27): Friske and Buriol found 2 new best known solutions to Group 1 instances: LR1_1_DR1_4_VC3_V11a and LR1_1_DR1_4_VC3_V12b for T=45 (i.e., a planning horizon of 45 periods).
- Update (2017-09-05): Papageorgiou et al. found new best known incumbents for 26 out of 70 yet-to-be-proved optimal Group 2 instances and new best known bounds on 56 of these same Group 2 instances.;
Please view the legend below for more details.
The instances can be partitioned into three main groups. Group 1 and Group 2 were presented in Papageorgiou et al. (2014) and can be distinguished by the length of the planning horizon. The Jiang-Grossmann (JG) instances were presented in Jiang and Grossmann (2015).
Group 1: Instances with a planning horizon of 60 periods or fewer having multiple ports per region and often requiring split pickups or split deliveries. Finding a single feasible solution is a challenge. A mixed-integer linear programming formulation is available here.
Group 2: Instances with a planning horizon greater than 60 periods having one port per region and never involving split pickups or split deliveries. On the other hand, instances in this group possessing many ports often involve travel times ranging from 5 to 37 periods. Finding a good feasible solution in minutes is a challenge. A mixed-integer linear programming formulation is available here.
Group 3: Jiang-Grossmann (please see Jiang and Grossmann (2015) for more details)
- Group 1: Group1_data_format_only_files, Group1_LP_files, Group1_MPS_files, Group1_Incumbents
- Group 2: Group2_data_format_only_files, Group2_LP_files, Group2_MPS_files, Group2_Incumbents
- Groups 1 and 2: Group1_and_2_data_format_only_files, Group1_and_2_LP_files, Group1_and_2_MPS_files
- Group 3: Jiang_Grossmann_GAMS_files, Jiang_Grossmann_LP_files, Jiang_Grossmann_MPS_files
- Data only format instructions