Based on version 0.3 of code located here.
The first MATLAB prototype of this code (not included here) was partially based on code developed by Andrea Censi and not available anymore on the web (at least I can't find it -- see references in the papers). The current release has been rewritten in C++ from scratch. This release implements the improvementes based on randomization in the Hough transform described in the IROS 2008 paper.
You are welcome to use this code according to the terms specified in the DISCLAIMER file. If the results you produce are used for scholar publications, please cite the following papers where the algorithm was described (basic version and improved version)
- S. Carpin. "Fast and accurate map merging for multi-robot systems". Autonomous Robots http://dx.doi.org/10.1007/s10514-008-9097-4
- S. Carpin. "S. Carpin. "Merging maps via Hough transform".Proceedings of the 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems
Online versions of these papers are available on here.
This code has been developed and tested on the following systems:
- macOS 10.5.4 with gcc 4.0.1
- Ubuntu 8.04 with gcc 4.2.3 (kernel 2.6.24-16-generic)
In order to build the code you need the opencv library. It is
assumed opencv was installed in /usr/local. If a different
path was used, just edit the Makefile
To build the library, just type "make".
Please note that the Makefile included in this version will build a shared library for Linux (i.e. a .so file). If you want to build it for macOS comment/uncomment the appropriate lines (see instructions in the Makefile)
Datasets used in the papers are available in the dataset folder. Each of them represents an occupancy grid map encoded according to the following: 0 is an occupied cell 127 is an unknown cell 255 is a free cell
Documentation for indidual functions and classes can be automatically generated using Doxygen. Note that this functionality was developed and tested only with Doxygen 1.5.6 so it may not work with other versions. To build the documentation just type "make doc".
See test.cpp for a simple example. In essence you need to setup two maps that should contain cells of three types: free, occupied and unwnown. If uncertain about the values to use, keep the defaults. After that, just call get_hypothesis or get_hypothesis_robust. That's it. In the current implementation the two maps must have the same size. This is not a intrinsic requirement of the algorithm, but rather a way to simplify the implementation. In a future version I will (perhaps "may" is a better word) remove this requirement.