graphhopper / jsprit

jsprit is a java based, open source toolkit for solving rich vehicle routing problems

jsprit

jsprit is a java based, open source toolkit for solving rich Traveling Salesman Problems(TSP) and Vehicle Routing Problems(VRP). It is lightweight, flexible and easy-to-use, and based on a single all-purpose meta-heuristic currently solving

• Capacitated VRP
• Multiple Depot VRP
• VRP with Time Windows
• VRP with Backhauls
• VRP with Pickups and Deliveries
• VRP with Heterogeneous Fleet
• Time-dependent VRP
• Traveling Salesman Problem
• Dial-a-Ride Problem
• Various combination of these types

Setting up the problem, defining additional constraints, modifying the algorithms and visualising the discovered solutions is as easy and handy as reading classical VRP instances to benchmark your algorithm. It is fit for change and extension due to a modular design and a comprehensive set of unit and integration-tests. More features ...

The jsprit-project is maintained by GraphHopper.

Getting Started with Documentation

Please visit docs to learn more.

Modules and Dependencies

Please read Notice.md to get to know the direct dependencies of each module.

Whats New

jsprit develops fast. Look here to get to know whats new.

License

This software is released under Apache License v2.

Contribution

Any contribution is welcome. Feel free to improve jsprit and make pull requests. If you want to contribute to jsprit (which would be great), fork the project and build your fork, make changes, run your and jsprit's test cases and make a pull request (see help.github.contribute or stackoverflow.contribute for details).

See who has contributed here.

Contact

Mailing List:

In the Graphhopper Forum you can discuss jsprit related issues and you will probably get answers to your questions.

Stackoverflow:

You can also use stackoverflow to discuss your issues. Tag it with jsprit then it is easier to keep track of your topic.

Issue Tracker:

For bugs, feature requests or similar use the issue tracker.

Email:

If you cannot get help in the mailing list or you just do not want to discuss your topic publicly, contact us via mail

jsprit's Issues

enhance SolutionPrinter to print out route stopps

add benchmark-wiki for multiple depot wiki with downloadable algorithm-config

test darp

code pd-algorithm

code local insertionCalc
code routeLevel insertionCalc

refactor vehicle-type from VehicleImpl

Add VRP with Backhauls (incl. documentation)

code darp

Issue solved

AlgoConfig: in prematureBreak the sequence of iterations and threshold should not matter

add and improve javadocs for basic classes

Document algorithm config and algorithm customizing

multiple depot vrp example is extremely slow since too many redundant vehicle-types have been defined

test rapi

add benchmark for capacitated vrp

Add pickup and delivery

this does not mean enroute pickup and delivery. Either the pickup or the delivery must take place in a depot

design rapi

design pd-problem

code problem

data and builders

Delivery-Points dependant on specific depot

Hi,

congrats for the awesome project, it looks really great!

I've got a feature request where there's a dependency between deliverys and depots, i.e. a multi-depot problem where a delivery can only be delivered from a specific depot.

Example:
Two depots, Depot A hosts Apples, B hosts Books.
Five Deliverys: 1 (gets Apples), 2 (Apples), 3 (Books), 4 (Apples + Books), 5 (Books)

Currently it's possible to solve the problem only if it doesn't matter from which depot the customers get their product, but if there are different products like in this example, it's more complicated. It would be great if something like the following would be possible:

Service.Builder.newInstance("1", 1).setCoord(Coordinate.newInstance(5, 7)).setDepot(...)

or

Service.Builder.newInstance("1", 1).setCoord(Coordinate.newInstance(5, 7)).setDepotCoord(...)

i.e. if one could link the service and the depot it gets deliverd from.

add benchmark for vrp with time windows

Add example reading and solving classical VRP instances

design darp

relax/code rapi

code additional builders and factories to decouple/hide implementation

speed up computation to better deal with large (cordeau-)instances (finite fleet and multiple depots)

better comp time can be achieved by:

• switching off time-window updates (since there are no time-windows)
• clustering customers (i.e. defining a neighborhood)
• improving vehicleFleetManager such that penaltyVehicles are only activated if needed (now they are always considered)

test pd-problem and algorithm

add prematureBreak configuration to algorithm config

extend analysis

e.g. to plot shipments

Add example capacitated vrp

Error when reading algorithm from file

reading an algorithm-file from any local folder returns error - cannot find file

Add example vrp with time windows

Add example time-dependent vrp

prematureBreak based on "variationCoefficient" in the algo-config file does not work in the SimpleExample

40000

0.001
2000

According to config, the iteration should stop after 2000 iterations.But it runs up 40000 iterations.

code shipment-io

problem readers/writer

add benchmark for vrp with backhauls

Add example comparing two algorithms

benchmark pd

dial a ride problem builder and solver

create benchmarker to benchmark algorithms

All example in wiki assume output folder - if it does not exist, create it automatically

Add example multiple depot vrp

design stable API to open up jsprit to create whole new algorithms from scratch

currently algorithms can only be customized via the algorithm-xml-file which is fine but it prevents users from building whole new algorithm (using availbable modules) with custom objective functions

design algorithm

here: ShipmentInsertionCalculator

Add example heterogeneous fleet

adjusts CalculateServiceInsertionOnRouteLevel the it was done with CalculateServiceInsertion

replace internal marginal cost calc with MarginalCalculus and add interface to route level costs estimates

design/code pd-constraints

add benchmark for vrp with pickup and deliveries

benchmark darp

enhance prematureBreak of algorithm based on percentage change

currently it is solely based on iterations

Create interface to MATSim

Add PenaltyVehicleType to better deal with penalty vehicles