Skip to content

Implementation of the A* / Dijkstra search algorithms, a Probabilistic Roadmap (PRM) Planner with path smoothing, and a statistical benchmarking suite

Notifications You must be signed in to change notification settings

nicholasRenninger/AStar_and_PRM_Planning_Analysis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

78 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Motion Planning - A* / Dijkstra and Probabilistic Roadmap (PRM) Planning

For optimal viewing of this document (and all *.md files), try opening it in a text editor that supports syntax highlighting for markdown *.md files (e.g. Sublime Text 2+).

Implementation of the A* and Dijkstra optimal graph search algorithms, a Fast Probabilistic Roadmap (PRM) Planner with path smoothing, and a benchmarking suite for doing parametric performance evaluation of the planners modules.

Built with a networkx Graph backbone and library implementations of:

  • Set-Priority Queue (based on heapq)
  • KDTree (based on cKDTree)
  • Union-Find (based on newtorkx.util)

ADTs for speed.

Nicholas Rennninger


SLOC Stats for nerds

-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Python                          26           1032           2192           2298
YAML                            18            206            404            443
Markdown                         1             21              0             43
-------------------------------------------------------------------------------
SUM:                            45           1259           2596           2784
-------------------------------------------------------------------------------

How to Run the Code

This repository uses python 3, which is managed with an anaconda environment.

Dependencies

In order to obtain all dependencies, you simply need to install anaconda for your OS, and then run from any shell with the conda command on its path:

$(REPO_DIR_LOCATION) $ conda env create -f conda/environment.yml

Running the Source Code

To run the code:

  • Activate the conda environment you just created in the "Dependencies" section: $(REPO_DIR_LOCATION) $ conda activate motionPlanning3

  • Run the main python module to run all of the simulations: $(REPO_DIR_LOCATION) $ python main.py


Configuring the Code

A* and Dijkstra Optimal Search Algorithms

To change the scenarios for the search algorithm simulations, simply modify one of the $(REPO_DIR_LOCATION)/config/graphSearch_XXX.yaml files to change the test graph or to change the location of the start / goal nodes.

PRM planner for a point Robot

To change the scenarios for the PRM planning simulation, simply modify one of the $(REPO_DIR_LOCATION)/config/prmPointRobot_XXX.yaml files to change the robot or obstacle vertices, or to change the location of the start / goal state or the robot, or to change properties of the PRM planner.

PRM planner Benchmarking

To change the scenarios for the PRM planner benchmarking, simply modify one of the $(REPO_DIR_LOCATION)/config/prmPointRobotBenchmark_XXX.yaml files to change the robot or obstacle vertices, or to change the location of the start / goal state or the robot, or to change properties of the PRM planner or to change the parameters you would like to vary.

Tool Class Hierarchy

This diagram shows the exact class design for the underlying python framework:


Results

A* and Dijkstra Optimal Search Algorithms

Graph 1 - A*

Graph 1 - Dijkstra

PRM Planner Viz and Benchmarking Stats

Environment 1 - PRM Viz

Environment 1 - PRM Benchmarking Results

Environment 2 - PRM Viz

Environment 2 - PRM Benchmarking Results

Environment 3 - PRM Viz

Environment 3 - PRM Benchmarking Results

About

Implementation of the A* / Dijkstra search algorithms, a Probabilistic Roadmap (PRM) Planner with path smoothing, and a statistical benchmarking suite

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages