Data for: Solving the Paint Shop Problem Using Reinforcement Learning
Description
This repository contains the instances of the paint shop problem for our paper "Solving the Paint Shop Problem Using Reinforcement Learning". The repository provides four datasets. 1.) The main dataset in the folder quadratic_buffer contains 170 problem instances for the paint shop problem with different buffer sizes. We sampled the incoming sequence from a multinomial distribution with equal probability for each color. 2.) The folder imbalanced_color_distributions contains 340 problem instances with imbalanced color distributions, where the probabilities decrease linearly or exponentially. 3.) The folder initially_filled_buffer contains 170 problem instances in which a buffer is sampled in addition to the incoming sequence. 4.) The folder rectangular_buffer contains 110 problem instances with rectangular buffers. Dataset structure Each folder contains several instance files. An instance file is a combination of the number of times each color occurs in the incoming sequence, the buffer size and the incoming sequence itself. File name notation: instance_A_B_C_D.mix A: Number of colors (5,10,15) B: Size of quadratic buffer (2-8) [For rectangular buffer: B = B1 x B2 with B1: number of lanes (4-10) and B2: length of each lane (4-10)] C: Length of incoming sequence (100) D: Instance number (0-9) Each file represents a different problem in text format with the following notation: Line 1: Number of times each color occurs in the incoming sequence Line 2: List of length of each buffer lane Line 3: Incoming sequence [For initially filled buffer: Remaining lines: Content of buffer lanes, one lane per line]