This repository contains the source code to build and solve a combinatorial lower bound model for the HHCRSP, introduced by Mankowska et al. (2014). The CLB model consists of a VRPTW without the constraints that set the time window ending. More precisely, the model comprises a multi-attribute VRP with constraints for the earlier visit times.
The working idea of such an approach is to solve a relaxed yet integer version of our problem of interest-the HHCRSP-so the solver can easily prove optimal solutions for such relaxed problem and, hopefully, producing strong bounds than the best lower bound from standard MIP-based solvers.
This project uses the VRPSolver 0.4.1 and runs thourgh a Docker container by design. To run an experiment, we suggest the use of VRPSolver script as the entry point of the application. The repository also contains the complete MIP model for the HHCRSP, and it also supports running the CLB model with CPLEX. Despite that, the user need to change the julia code to change which solution method should be employed to solve the CLB model.
To run the software, you need to follow the install instructions according VRPSolver documentation. Note that you will also need a supported version of CPLEX. VRPSolver documentation has more information about that.
As Docker operates in isolated containers, the most straightforward option to read and write data from it is through mapped mounting points. The user does not need to worry about this technicality, as the VRPSolver script abstracts these extra configuration steps. This way, the only step necessary prior running the code is copy the instances file to the data directory. For now, the solver only outputs data to stdout and there is no such a output directory.