MaxMagazin / ma_rrt_path_plan

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

RRT with multiple remote goals.

This repository contains the ROS package with code implementation of RRT-based path planning algorithm suitable for exploration of a trackdrive circuit according to the rules of Formula Student Driverless competition in Germany. This algorithm is based on my contribution to E-gnition, a FS Team of Technical University Hamburg.

Basic animation of the approach

A brief introduction to the main steps of the proposed algorithm is given in my Master's thesis presentation (direct timestamp).

Notes

  • The algorithm does not utilize the cones' color information, but instead a simple logic is used, which rewards the branches with cones from both sides (see findBestBranch(...) method), and penalizes the branches having cones only from one side. With cone classification information a better rating system can be implemented and applied.
  • Unfortunately I wasn't able to test and see this algorithm working on real hardware, a FS Driverless car, so I am excited if you can bring it the reality on any mobile robot and share some videos with me (see section Usage)

Usage

Inputs, outputs, params

Inputs

  • rospy.Subscriber("/map", Map, ...)
  • rospy.Subscriber("/odometry", Odometry, ...)

Outputs

  • rospy.Publisher("/waypoints", WaypointsArray, ...)

Outputs (Visuals)

  • rospy.Publisher("/visual/tree_marker_array", MarkerArray, ...)
  • rospy.Publisher("/visual/best_tree_branch", Marker, ...)
  • rospy.Publisher("/visual/filtered_tree_branch", Marker, ...)
  • rospy.Publisher("/visual/delaunay_lines", Marker, ...)
  • rospy.Publisher("/visual/waypoints", MarkerArray, ...)

Parameters to tune (main)

  • planDistance = 12 m - Maximum length of tree branch
  • expandDistance = 1 m - Length of tree node/step
  • expandAngle = 20 deg - constraining angle for next tree nodes
  • coneObstacleSize = 1.1 m - size of obstacles derived from cone position
  • coneTargetsDistRatio = 0.5 - ratio to frontConesDist for deriving remote cone goals

Licence

MIT

  • Use me and my code in any way you want
  • Keep the names and the same licence

About

License:MIT License


Languages

Language:Python 95.7%Language:CMake 4.3%