GPPC: Grid-Based Path Planning Competition
------------------------------------------------------------ CALL FOR PARTICIPATION: GRID-BASED PATH PLANNING COMPETITION Classic Track 2014 ------------------------------------------------------------ A wide variety of researchers have used grid-based maps as testbeds for search and they are also used in many deployed applications. There has been, however, a shortage of comparisons between existing work, and many published techniques may not apply in more real-world situations. The advent of widely available benchmark problems (movingai.com/benchmarks) is one step towards encouraging comparison between existing approaches. The third annual grid-based path planning competition will be held in 2014, with the goal of furthering comparisons between existing work and encouraging the publication of code for fast pathfinding implementations. Teams are required to register for the competition by midnight December 1 (UTC-12). Entries must be finalized by midnight December 15 (UTC-12). But, entries can be submitted at any time prior to this. TECHNICAL DETAILS ----------------- The competition will be run with the following constraints: * Maps will be at most 2048x2048. * The maps will be static (unchanging). * Maps will be 8-connected; in order to move diagonally, both adjacent cardinal directions must be unblocked. * Diagonal actions in an optimal path cost sqrt(2). * The competition will be broken into pre-processing and testing phases. ** In the pre-processing phase, entries will pre-process a map, writing any necessary files to disk. ** In the testing phase, entries can load the pre-processed data from disk to use for path planning. ** Pre-processing time and disk limits have been lifted, but the use of significant resources may delay your results. * All entries will be made available as open source after the competition. * Programs must be written in C/C++ or be written to link to sample code from which the tournament will be run. Sample code for the competition is available from: http://www.movingai.com/GPPC/code.html The sample code performs three main calls to the code provided in a tournament entry: void PreprocessMap(std::vector&bits, int width, int height, const char *filename); void *PrepareForSearch(std::vector &bits, int width, int height, const char *filename); bool GetPath(void *data, xyLoc s, xyLoc g, std::vector &path); The main.cpp sample file should not be modified as it may be replaced for the actual competition. Each entry should include a makefile for compiling the code for the competition. If participants wish to enter using a different language besides C/C++, they should contact the tournament organizer . WINNER DETERMINATION -------------------- Instead of competing on a particular metric, metrics will be computed about the performance of each entry in the competition. Dominated entries (which are worse than some other program in all metrics) will be noted, but the remaining entries are all considered viable approaches depending on the metrics required for a particular domain. The metrics tested will include: * The total time to find a complete path. * The time to find the first 20 steps of the path. * The maximum time for returning any portion of a path. * The total path length (suboptimality). * The total memory (RAM) usage. * The total disk usage. * The pre-processing time. SUBMISSION INFORMATION ---------------------- To register for the competition, please fill out the entry form at: http://movingai.com/GPPC/register.html PREVIOUS ENTRIES ---------------- Previous GPPC entries are available at: http://movingai.com/GPPC/ GPPC Chair: Nathan Sturtevant, University of Denver sturtevant@cs.du.edu