VRPenstein is a parameterized meta-heuristic for the vehicle routing problems (VRP). Currently, the configuration space of VRPenstein includes two different types of meta-heuristics:
- Memetic Algorithm (Population-based search with local search)
- Iterated Local Search (Single point search)
Now the supported VRP variants include
- vehicle routing problem (VRP)
- capacitated vehicle routing problem (CVRP)
- capacitated vehicle routing problem with time windows (VRPTW)
- capacitated vehicle routing problem with pick-up and time windows (VRPPTW)
- simaultanes Delivery and Pick-up Vehicle Routing Problem with Time Windows (VRPSDPTW)
git clone [email protected]:senshineL/VRPenstein.git
mkdir Bin
cd SRC
g++ -std=c++11 -o ../bin/VRPenstein -O3 MemeticAlgorithm.cpp eval.cpp operator.cpp search_framework.cpp solution.cpp util.cpp data.cpp
VRPenstein --problem PROBLEM [--pruning] [--output OUTPUT] [--time TIME] [--runs RUNS] [--g_1 G_1] [--pop_size POP_SIZE] [--init INIT] [--k_init K_INIT] [--cross_repair CROSS_REPAIR] [--k_crossover K_CROSSOVER] [--parent_selection PARENT_SELECTION] [--replacement REPLACEMENT] [--ls_prob LS_PROB] [--O_1_eval] [--two_opt] [--two_opt_star] [--or_opt OR_OPT] [--two_exchange TWO_EXCHANGE] [--elo ELO] [--random_removal] [--related_removal] [--alpha ALPHA] [--removal_lower REMOVAL_LOWER] [--removal_upper REMOVAL_UPPER] [--regret_insertion] [--greedy_insertion] [--rd_removal_insertion] [--bks BKS] [--random_seed RANDOM_SEED]
The MATE Algorithm [http://arxiv.org/abs/2011.06331]
The MATE algorithm is a special case of VRPenstein. To repeat our experiments:
VRPenstein [--problem PROBLEM] --pruning [--time TIME] --g_1 50 --[pop_size POP_SIZE] --init rcrs --cross_repair regret --parent_selection circle --replacement one_on_one --O_1_eval --two_opt --two_opt_star --or_opt 2 --two_exchange 2 --elo 1 --related_removal --removal_lower 0.25 --removal_ipper 0.40 --regret_insertion
All the benchmarks are in data/data.tar.gz
This software is released under MIT license.