https://github.com/realkushagrakhare/3d_path_planning
AI project for 3D Path Planning. Other details and running instructions can be found on the Readme.md file
https://github.com/realkushagrakhare/3d_path_planning
3d-path-planning 3d-pathfinding artificial-intelligence artificial-intelligence-algorithms collision-detection constraint-satisfaction-problem constraints dubins dubins-rrt motion node-prune path-planning planning-algorithms rapidly-exploring-random-tree reeds-shepp-planner rrt rrt-star satisfaction spline-fit spline-interpolation
Last synced: 3 months ago
JSON representation
AI project for 3D Path Planning. Other details and running instructions can be found on the Readme.md file
- Host: GitHub
- URL: https://github.com/realkushagrakhare/3d_path_planning
- Owner: realkushagrakhare
- License: mit
- Created: 2018-08-27T06:10:15.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2018-12-07T11:14:40.000Z (over 6 years ago)
- Last Synced: 2025-01-31T04:16:32.502Z (4 months ago)
- Topics: 3d-path-planning, 3d-pathfinding, artificial-intelligence, artificial-intelligence-algorithms, collision-detection, constraint-satisfaction-problem, constraints, dubins, dubins-rrt, motion, node-prune, path-planning, planning-algorithms, rapidly-exploring-random-tree, reeds-shepp-planner, rrt, rrt-star, satisfaction, spline-fit, spline-interpolation
- Language: Python
- Homepage:
- Size: 5.74 MB
- Stars: 115
- Watchers: 2
- Forks: 38
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 3D_Path_Planning
### is an AI project for 3D Path Planning which involves pruning with constant satisfaction.
3D path planning is required in various applications such as robotics, self-driving cars,
protein folding, games etc. It ensures to and a trajectory from the initial point to the
destination, subject to rules of motion and any other constraints, such as collision avoid-
ance, balance and joint limits.
Algorithms like Dijkstra, A* can be used but they are quite expensive to compute for
large clustered space Random sampling based planning algorithm like RRT can solve
motion planning problem while also taking the differential constraint into consideration.
But the paths so produced are jagged, with several unnecessary branches. They need
to be pruned and smoothed. An approach could be to fit a spline over the points which
would produce a smooth path.
Developers: Kushagra Khare, Rachit Jain
Mentor: Prof. Srisha Rao, IIIT-B
Project Duration: Aug '18 - Nov '18
## Goal
In this project, we aim to provide an algorithm for 3D Path Planning. We will implement a
RRT-A* based 3D Path Planning algorithm. The algorithm would include path pruning with
constraint satisfaction and account for non-holonomic constraints. We will go ahead with
Manhattan based RRT-A* in the initial stages but will also try to find an optimized distance
metric function using Voronoi bias property for the algorithm.
## Technologies used:
- PyGame & POGL
- Scikit, NumPy, Scipy
## Running Requirements
- Python 3.7
- Pygame and Scipy
## Milestones
### Week 1
- Implemented a basic Random-exploring Random Tree algorithm using PyGame.
- Read about various types of RRTs and implemented RRT-A* vartion. Compared its pros and
cons with the basic algorithm
### Week 2
- Implemented RRT* which results into asymptotically optimum solution.
### Week 4
- Added Node Pruning and Spline Fitting in RRT* pipeline.
- Later we realized that node pruning is not needed for RRT* as probability of pruning a path of RRT* is nearly 0.04.
### Week 5
- Implemented 2-Phase Sampling making the algorithm faster.
### Week 7
- Combined Dubins and Reeds-Shepp Path Planning algorithms with RRT* for non-holonomic constraints.
## Future works:
- Obstacle collision detection can be parallelized by using CUDA which will make computation a lot faster and applicable in real life scenarios.
- Including various other non-holonomic constraints like velocity, size of robot(or car).