-
Notifications
You must be signed in to change notification settings - Fork 2
Home
Welcome to the wiki of the multi-robot path planning (MRPP) project!
Although there are numerous theoretical approaches for the multi-agent pathfinding (MAPF) problem, there are only a few implemented and tested on robots. The purpose of this project is to bridge this gap through researching, implementing, and evaluating different algorithms that allow multi-robot path planning and collision avoidance.
As a first milestone, the execution and evaluation are simulated only. In future work, some of the developed programs are supposed to be executed on real robots. Additionally to the motion planning algorithm, the topology for information processing as well as communication can be chosen freely. The localization is done using the "adaptive Monte Carlo localization" (AMCL) approach by means of odometry.
The first step was to conduct thorough research on existing multi-agent and multi-robot algorithms as well as possible map representations. The results were systematized and summarized in research results section.
During that time, every member decided on an algorithm which mainly determined the further course of actions, since we developed our code separately. Though, in regular meetings, we ensured an exchange of knowledge and provided support whenever possible.
Basically, we came up with two different approaches in terms of coding:
- migrating and enhancing an existing ROS package including ROS move base,
- Coding of the path planning algorithm from scratch (without ROS move base).
A big hurdle for all of us was lacking knowledge about ROS and its components. That is why we want to provide some help for the future teams (see Helpful links for ROS).