lkh 2.0.0

Super simple Python wrapper for LKH-3

PyLKH

This is a super simple Python wrapper for the constrained traveling salesman and vehicle routing problem solver called LKH-3.

If you want to use this wrapper, you need to install LKH-3 first. For example, on Ubuntu:

wget http://akira.ruc.dk/~keld/research/LKH-3/LKH-3.0.6.tgz
tar xvfz LKH-3.0.6.tgz
cd LKH-3.0.6
make
sudo cp LKH /usr/local/bin

LKH-3 expects problems in the TSPLIB format. It extends the format to support VRPs.

Using PyLKH you can solve problems represented as Python objects or files.

CAUTION: distances are represented by integer values in the TSPLIB format. This can produce unexpected behaviour for some problems, like those with all nodes within the unit square. You can use the EXACT_2D distance to avoid rounding issues.

Install

pip install lkh

Example

import requests
import lkh

problem_str = requests.get('http://vrp.atd-lab.inf.puc-rio.br/media/com_vrp/instances/A/A-n32-k5.vrp').text
problem = lkh.LKHProblem.parse(problem_str)

solver_path = '../LKH-3.0.6/LKH'
lkh.solve(solver_path, problem=problem, max_trials=10000, runs=10)

Output (values correspond to nodes, which are 1-indexed, not node indicies, which are 0-indexed):

[[27, 8, 14, 18, 20, 32, 22],
 [25, 28],
 [15, 29, 12, 5, 24, 4, 3, 7],
 [30, 19, 9, 10, 23, 16, 11, 26, 6, 21],
 [13, 2, 17, 31]]

API

lkh.solve(solver='LKH', problem=None, problem_file=None, **kwargs)

Solve a problem instance.

Parameters

• solver (optional): Path to LKH-3 executable. Defaults to LKH.

• problem (optional): Problem object. LKHProblem is preferred but tsplib95.models.StandardProblem also works. problem or problem_file is required.

• problem_file (optional): Path to TSPLIB-formatted problem. problem or problem_file is required.

• kwargs (optional): Any LKH-3 parameter described here (pg. 5-7) or here (pg. 6-8). Lowercase works. For example: runs=10.

Returns

routes: List of lists of nodes (nodes, not node indicies).

class lkh.LKHProblem

A problem that can be solved by LKH-3. Fields are (partially) described here (pg. 4-6). Inherits from tsplib95.models.StandardProblem.

The available specification fields are:

• CAPACITY
• COMMENT
• DEMAND_DIMENSION
• DIMENSION
• DISPLAY_DATA_TYPE
• DISTANCE
• EDGE_DATA_FORMAT
• EDGE_WEIGHT_FORMAT
• EDGE_WEIGHT_TYPE
• NAME
• NODE_COORD_TYPE
• RISK_THRESHOLD
• SALESMEN
• SCALE
• SERVICE_TIME
• TYPE
• VEHICLES

The available data fields are:

• BACKHAUL_SECTION
• CTSP_SET_SECTION
• DEMAND_SECTION
• DEPOT_SECTION
• DISPLAY_DATA_SECTION
• DRAFT_LIMIT_SECTION
• EDGE_DATA_SECTION
• EDGE_WEIGHT_SECTION
• FIXED_EDGES_SECTION
• NODE_COORD_SECTION
• PICKUP_AND_DELIVERY_SECTION
• REQUIRED_NODES_SECTION
• SERVICE_TIME_SECTION
• TIME_WINDOW_SECTION

You probably want to initialize a problem instance using one of the following class methods:

classmethod load(filepath, **options)

Load a problem instance from a text file.

Inherited from tsplib95.problems.Problem.load.

classmethod parse(text, **options)

Parse text into a problem instance.

Inherited from tsplib95.problems.Problem.parse.

classmethod read(fp, **options)

Read a problem instance from a file-like object.

Inherited from tsplib95.problems.Problem.read.