Towards Intelligent Navigation and Collision Avoidance for Autonomous 3D Exploration with Dynamic Obstacles
Image: Gränsö Village in Gazebo. Used for evaluating the planner in a large-scale environment with multiple pedestrians. This area is roughly 1 hectare.
This repository is linked to Ludvig and Emil's Master Thesis Work at Linköping University.
This master's thesis aims to investigate how a system can be built to autonomously explore and map a 3D environment where dynamic obstacles are present.
It resulted in a modified and extended version of the Autonomous Exploration Planner AEP, called Dynamic Autonomous Exploration Planner (DAEP).
To evaluate DAEP, the planner was compared with AEP, DEP, and RH-NBVP. This repo contains functionality to run the mentioned planners in a rootless docker environment.
For more information on how daeplanner works, please see the thesis report.
The system was tested on a computer with:
- AMD Ryzen 9 5900X 12-core processor
- NVIDIA GeForce RTX 3090 Ti
- Ubuntu 22.04.2 operating system.
- Rootless docker mode.
See Wiki.
These GIFS present the exploration with DAEP in two different scenarios. A video is also available here.
Exploration in the Cafe World with the DAEP. White cylinders represent dynamic obstacles. The green area represents the the future uncertainty of pedestrian motion.
Exploration in the Crosswalks World with DAEP. White cylinders represent dynamic obstacles. The green area represent the the future uncertainty of pedestrian motion.
This work has been developed by Ludvig Widén and Emil Wiman with the help and support of the members at the Division of Artificial Intelligence and Integrated Computer Systems at Linköping University.
For questions and remarks please contact the developers by supplying an issue.