Overview

Mind map which describes in short project's goals, used algorithms, and challenges.

solvermission/goalbest feature setas many features as possibleout of the boxgood qualityclose to best knownfastreturn acceptable solutions fastlow resource consumptionmemorycpufeaturesvariantsCapacitated VRP (CVRP)Heterogeneous Fleet VRP (HFVRP)VRP with Time Windows (VRPTW)VRP with Pickup and Delivery (VRPPD)VRP with backhauls (VRPB)Multi-Depot VRP (MDVRP)Multi-Trip VRP (MTVRP)Multi-Objective VRP (MOVRP)Open VRP (OVRP)VRP with Lunch Break (VRPLB)VRP with Route Balance (VRPRB)Periodic VRP (PVRP)Time dependent VRP (TDVRP)Skill VRP (SVRP)Traveling Salesman Problem (TSP)...informalstablepickups, deliveries, skills, etc.multi-location jobinitial solutionscientific formats...experimentaljob typeservicereplacementmulti-jobminor perf. improv.vehicle placedispatchreloadbreaklegal breakmultiple breaksunassigned break weightmulti tourtour balancingtime dependent routingmultiple solutions...heuristicsconstructiveinsertioncheapestn-regretskip n-bestblinks+ 3 morenearest-neighbourmetamutationruin recreate (LNS)ruinadjusted string removal (SISR)cluster removal (DBSCAN)random job removal+ 3 morerecreatereuse constructuve heuristicslocal searchinter route exch.intra route exch.decompositiondecompose solution into smaller onescreate and solve smaller problems independentlycompose a new solution from partial onesdiversificationrosomaxacluster solutions by ANN (GSOM)2D search process visualizationdiversity tuningelitegreedyobjectivekindmulti (NSGA-II)hierarchicaltypesminimize/maximize routesminimize costminimize unassignedtour balancinghyperkindselectionfixed probabilitiesselect from the listcombine multipledynamic probabilitiesMDP modelapply RLgenerativeTBDchallengesexploration/exploitation dilemmaissuesstagnationunstable quality resultssolutionsimprove meta-heuristicmore ruin/recreatesoptimal deconstruction (removal) sizemore local search operators (e.g. 2-opt.)extra mutation typesimprove hyper-heuristicRL/MDP: dynamic probabilities [WIP]ROSOMAXA: dynamic parametersalgorithm optimizationsdata parallelism controlcachingfeature requirementsissuesalgorithm extensibilityinsertion heuristic assumptionsruin/recreate approachfeature interferencecommon format representation