Trace explosive detection method based on multi-UAV formation

Duration: Aug. 2020 - Nov. 2020

Advisor: Xiaoduo LI

Background

Much attention has been paid to the detection of explosive compounds in antiterrorism applications. Facing the increasing demand for trace explosive detection, the company plans to employ UAVs equipped with explosive detection device to realize the detection of dangerous goods in a larger range with high crowd density. In order to initially verify the feasibility of the idea, it is planned to use Gazebo virtual simulation software to construct a UAV trace explosive detection scenario, where multi-UAV systems aim to realize a time-varying formation to cover a larger detection range.

Intership content

Specific internship content includes three parts:

First, learn to use Gazebo virtual simulation software, and construct a trace explosive detection scenario in the Gazebo world.

Second, utilize the reinforcement learning method based on dynamic Q-learning to train the fixed-point tracking algorithm of a single UAV control.

Third, design a multi-UAV intelligent formation control method based on the reinforcement learning algorithm, so as to realize the objective of multi-UAV trace explosive detection in a large space.

Achievements

1. This photo displays the scenario of trace explosive detection in Gazebo world. 👇

   

2. The following video shows the training process of using reinforcement learning algorithms to control a single UAV.

3. This video indicates the scenario of multi-UAV systems formation tracking with four followers to realize the trace explosive detection. 😎