Overview

(the picture is taken from en.wikipedia.org/D*)

dstar_trajectory_planner is a move_base plugin implementing global trajectory planning with D* informed incremental search algorithm with external optimizers. It strives to simplify the development of robotic systems operating in dense industrial environments.

• Package URL: https://github.com/ElettraSciComp/DStar-Trajectory-Planner

Package Contents

The package contains two plugins:

• D-star trajectory planner that maintains generation of the global trajectory on the provided costmap, and dynamic replanning by request coming from move_base.

• Virtual Walls module processing JSON objects to generate the non-passable and explicitly passable zones as OpenCV polygons added over the existing map.

Configuration

Example Configuration

DStarGlobalPlanner:
    goal_distance_threshold: 0.3
    neighbor_distance_threshold: 0.12
    occupancy_threshold: 64
    cutoff_distance: 16
    trajectory_optimizer:
        repulsion_gain: 70.0
        potential_field_radius: 12
    erosion:
        enable: true
        erosion_gap: 2
    paths:
        enable: true
        from: map.json # OPTIONAL PASSABLE ZONES FILE

Parameters Explained

Development and Roadmap

The package was developed under the Remotization and Robotization Initiative at Elettra Sincrotrone Trieste by the members of the ICT Group. It is part of an innovative flexible control system for mobile robots that the group leaders presented in Tokyo during 16th IFToMM World Congress 2023. The interested ones can find the published papers here:

• Universal Configurable Navigation and Control System for Industrial Unmanned Ground Vehicles with Differential Chassis

• D-Star-Based Optimized Trajectory Planner for Mobile Robots Operating in Dense Environments

Roadmap

1. Improve the initial documentation.
2. Implement the custom "magic" linked list that keeps the sorted state.
3. Avoid map copying during the initialization.
4. Separate area parsers completely from the kernel library.