A Deep-Learning-based Strategy for Kidnapped Robot Problem in Similar Indoor Environment
- PDF / 1,317,344 Bytes
- 11 Pages / 595.224 x 790.955 pts Page_size
- 1 Downloads / 176 Views
A Deep-Learning-based Strategy for Kidnapped Robot Problem in Similar Indoor Environment Shikuan Yu1 · Fei Yan1
· Yan Zhuang1 · Dongbing Gu2
Received: 14 July 2019 / Accepted: 25 May 2020 © Springer Nature B.V. 2020
Abstract We present a deep-learning-based strategy that only uses a 2D LiDAR sensor to solve the kidnapped robot problem in similar indoor environments. First, we converted a set of 2D laser data into an RGB-image and an occupancy grid map and stacked them into a multi-channel image. Then, a neural network structure with five convolutional layers and four fully connected layers was designed to regress the 3-DOF robot pose. Finally, the network was trained using multi-channel images as input. We also improved the network structure to identify the scene where the robot is localized. Extensive experiments have been conducted in practice with a real mobile robot, verifying the effectiveness of the proposed strategy. Our network can obtain approximately 2m and 5◦ accuracy indoors, and the scene classification accuracy of our network reaches up to 98%. Keywords Relocalization · 2D LiDAR sensor · CNN · Robot pose · Kidnapped robot problem
1 Introduction Over the past few decades, indoor robotics has made an amazing improvement. The localization and navigation of mobile robots had attracted the attention of many scholars [1, 2]. Indoor robots can perform navigation tasks This work was supported by the National Natural Science Foundation of China (U1913201, 61503056) and the Science and Technology Foundation of Liaoning Province of China (20180520031). Fei Yan
[email protected] Shikuan Yu [email protected] Yan Zhuang [email protected] Dongbing Gu [email protected] 1
Key Laboratory of Intelligent Control and Optimization for Industrial Equipment of Ministry of Education and the School of Control Science and Engineering, Dalian University of Technology, Dalian, 116024, China
2
School of Computer Science and Electronic Engineering, University of Essex, Colchester, UK
on condition that they know their location. Most of the proposed strategies for robot localization are based on probabilistic estimation techniques [3, 4]. Particle Filter algorithm and its improvements are used in [5–8]. These algorithms have excellent performance in the field of robot localization, but all of them cannot solve the Kidnapped Robot Problem. The Kidnapped Robot Problem is still a major challenge in the field of indoor roboticsis, it is a localization problem which means a robot can not calculate it’s current pose, and this kind of problem often occurs when a robot wakes up in similar indoor environments. In recent years, approaches are aiming at solving the kidnapped robot problem for indoor environments. A method based on the adaptation of 2D Speeded Up Robust Features (SURF) image features to 3D landmarks of the environment was presented in [9]. This method searches for landmarks when the robot wakes up in an unknown area. This method built up a prior map containing 3D Surf landmarks, which can reduce the similarity
Data Loading...
