Probabilistic roadmap (PRM) algorithm generates graphs to perform path planning with complex constraints and high dimensions but has some limitations in situations like narrow pathways and environments with dynamic obstacles. This drawback of PRM is usually solved by increasing the number of randomly generated sampling points. However, too many sampling points will increase the computational complexity resulting in poor performance. Hence to overcome these limitations PRM with potential fields can be implemented. Potential field can be generated for the workspace, will help determine the complexity of workspace and adequate number of sampling points required, and then ensure high density of sampling points around the obstacles by implementing a regional sampling strategy.
- Calculate Repulsive potential for each point q in the map generated by obstacle qo.
- Determine Number of Sampling Points.
- Using bounding range for open area and obstacle region sampling point can be distributed in the map.
- Ubuntu18
- ROS Melodic
- python3.x
- NumPy
- cv2
- panda
- Matplotlib
- sklearn
- shapely
cd <your_workspace>/src
git clone https://github.com/Prat33k-dev/PRM-Blended-Potential-Field-Path-Planning.git
cd ../
rosdep install --from-paths src --ignore-src -r -y
catkin_make
First need to genrate .csv file of path coordinates to follow from start to goal location.
cd <your_workspace>/src/PRM-Blended-Potential-Field-Path-Planning/prm_apf_planner
python3 src/main.py --start 6 1 --goal 19 19 --FilePath './map/map.png'
- start - Start position of the robot. Default :- [6 1]
- goal - Goal position of the robot. Default :- [19 19]
- FilePath - map file path. Default :- './map/map.png'
cd <your_workspace>
source /devel/setup.bash
roslaunch prm_apf_planner simulation.launch
Environment | Occupancy Map |
---|---|
![]() |
![]() |
Around the Obstacle | Heatmap |
---|---|
![]() |
![]() |
[1] H. You, G. Chen, Q. Jia and Z. Huang, "Path Planning for Robot in Multi- dimensional Environment Based on Dynamic PRM Blended Potential Field," 2021 IEEE 5th Information Technology,Networking,Electronic and Automation Control Conference (ITNEC), 2021, pp. 1157-1162, doi: 10.1109/ITNEC52019.2021.9586848. (Link)
[2] P. Fankhauser and M. Hutter, "A Universal Grid Map Library: Implementation and Use Case for Rough Terrain Navigation", in Robot Operating System (ROS) – The Complete Reference (Volume 1), A. Koubaa (Ed.), Springer, 2016. (Link)
[3] https://www.cs.cmu.edu/~motionplanning/lecture/Chap4-Potential-Field_howie.pdf.