/EECE597

Trajectory planning with real time vision-based obstacle detection

Primary LanguageTeX

Trajectory planning with real time vision-based obstacle detection

This project implemented object detected with RealSense camera with RRT trajectory plan algorithm, simulated in MATLAB environment. This project is done during UBC project course EECE597. Thanks for the supervision of Dr. Maryam Kamgarpour and Dr. Mahdi Yousefi.

Introduction to the repo

There are three main parts for this project: trajectory plan algorithm, obstacle dection with RGBD image processing, and integration of the detected obstacle to trajectory plan.

The final report is here.

Further instruction on UBC report template is here.

Usage instruction

Packages need to be installed in MATLAB (Please contact me if other toolboxes needed):

  • ROS Toolbox

Run path planning with pre-defined objects

  • Clone the whole repo to local dir
  • Nevigate to folder PathPlanRRT in MATLAB. This can be done by opening one of the matlab scripts here, for exmaple pathPlan_main.m
  • Type this command in command window: run("pathPlan_main.m")
  • Add any path needed to be included

Run the object detection

  • Nevigate to folder ObjectDetection in MATLAB. This can be done by opening one of the matlab scripts here, for exmaple videoProcess_main.m
  • Type this command in command window: run("videoProcess_main.m")
  • Add any path needed to be included
  • Note: default path to the pre-saved Intel RealSense video is EECE597/Videos/, please change them accordingly in videoProcess_main.m

Run path planning with obeject detection integrated:

  • Nevigate to folder Integrate in MATLAB. This can be done by opening one of the matlab scripts here, for example integrate_main.m
  • Type this command in command window: run("integrate_main.m")
  • Add any path needed to be included
  • Note: to run the integrated result, user has to process the recorded video first with run("videoProcess_main.m") in folder ObjectDetection

Progress logs

1. Path planning with RRT, MATLAB (Feb 26, 2020)

2. Real-time path planning with RRT, add obstacle half way (March 4, 2020)

Acutal path

Another Example

3. Object Detection (March 11, 2020)

4. Improvement on Object Detection (March 25, 2020)

Top View image from depth info

5. Map the obstacle on path plan (Apr 1, 2020)

6. Map the obstacle with an ellipse boundary (Apr 9, 2020)

7. Update the path as obstacle moving (Apr 16, 2020)

8. Update the path as move obstacles detected (Apr 23, 2020)