|
Abstract
|
|
|---|---|
| In this paper, a fully-autonomous aerial robotic solution, for executing Search and Rescue (SAR) missions in unstructured indoor environments, has been developed. The proposed system is based on the combination of a complete hardware configuration and a flexible system architecture which permits the execution of high-level missions in a fully unsupervised manner. In order to obtain complex and robust behaviors from the proposed aerial robot, several deep learning capabilities have been integrated for target recognition and interaction. The target recognition capability is based on a supervised learning classifier developed as a Convolutional Neural Network (CNN) trained for target/background classification, while the target interaction capability introduces an Image-Based Visual Servoing (IBVS) algorithm which integrates a recent Deep Reinforcement Learning method named Deep Deterministic Policy Gradients (DDPG). In order to train the UAV for performing IBVS tasks, our own reinforcement learning framework has been developed, which integrates deep reinforcement learning capabilities (e.g. DDPG) with a Gazebo-based environment for aerial robotics. The proposed system has been validated in a wide range of simulation flights, using Gazebo and PX4 Software-In-The-Loop, and real flights in unstructured indoor environments, demonstrating the versatility of the proposed system. | |
|
International
|
Si |
|
JCR
|
Si |
|
Title
|
Journal of Intelligent & Robotic Systems |
|
ISBN
|
0921-0296 |
|
Impact factor JCR
|
1,512 |
|
Impact info
|
Datos JCR del año 2017 |
|
Volume
|
|
|
|
|
|
Journal number
|
|
|
From page
|
|
|
To page
|
|
|
Month
|
SIN MES |
|
Ranking
|
|