第 39 卷第 3 期 2017 年 5 月 机器人 ROBOT Vol.39, No.3 May, 2017 DOI：10.13973/j.cnki.robot.2017.0600 Indoor Positioning System Using Axis Alignment and Complementary IMUs for Robot Localization ZHOU Qinqin 1 , LEI Siyu 1 , YU Zhangguo 1 , LIN Hsien-I 2 , CHEN Xuechao 1 , ZHANG Weimin 1 (1. Intelligent Robotics Institute, Beijing Institute of Technology, Beijing 100081, China; 2. Graduate Institute of Automation Technology, Taipei University of Technology, Taipei 10608, China) Abstract: An indoor positioning method for robots is presented to improve the precision of displacement measurement using only low-cost inertial measurement units (IMUs). Firstly, a high-fidelity displacement estimation for linear motion is proposed. A new robot motion model is designed as well as an axis alignment that only uses a single axis of the accelerometer. The integral error of velocity is eliminated by a new subsection calculation method. Two complementary IMUs are combined by assigning them different weights to obtain high accuracy displacement results. Secondly, an orientation estimation based on a fusion filter for the steering motion is proposed. Experiments show that the proposed method significantly improves the accuracy of linear motion measurement and is effective for the indoor positioning of a robot. Keywords: axis alignment; displacement estimation; indoor positioning; orientation estimation; inertial measurement unit 1 Introduction Current technologies enable autonomous robot movement [1] . Autonomous movements require various sensors to detect environmental conditions as well as in- ternal conditions, and the obtained information is then processed to determine the subsequent movement. In addition to the obstacles and terrain common to an in- door environment, a robot must also be able to nego- tiate unpredictable and complex situations. Therefore, an accurate indoor positioning technique is essential to realize autonomous robot movement. Common positioning methods are divided into ab- solute positioning and relative positioning. These tech- nologies include: satellite positioning [2] , visual posi- tioning [3] and inertial positioning [4] . Satellite positioning technology, such as global po- sitioning system (GPS), is currently the most widely used method. This method comprises space satellites, ground control segments, and end-user equipment [5] . Although GPS provides high precision 3-dimensional positioning information, the signal is easily blocked by barriers, especially for indoor-based robots. Sig- nal blocking decreases positioning accuracy. Therefore, GPS is not an optimal choice for indoor positioning. Visual positioning technology commonly affixes cameras onto the robot to capture images and then pro- cesses these images using machine vision and other technologies to achieve positioning [6] . Although the equipment required for this technology is simple and inexpensive, the technology is limited by memory stor- age and processor power. Also, delays occur under the heavy processing workloads that are required to achieve positioning. Furthermore, on-board cameras are sensi- tive to light intensity and illumination while light may be insufficient or even absent for indoor environments. Inertial positioning technology mainly uses an in- ertial measurement unit (IMU) to obtain information and then integrates acceleration according to Newton’s second law to achieve indoor positioning [7] . There- fore inertial positioning technology is not affected by barriers or inhibited by lack of lighting; also, the com- putational overhead is low. Compared with the other two positioning technologies, inertial positioning tech- nology is more suitable for indoor environments [4] . Chen L-H et al. proposed a maximum likelihood- based fusion algorithm that integrated a typical Wi-Fi indoor positioning system with a pedestrian dead reck- Supported by: National Natural Science Foundation of China (61375103, 61533004, 61320106012, and 61321002); the 863 Program of China (2014AA041602, 2015AA042305 and 2015AA043202); the Key Technologies Research and Development Program (2015BAF13B01 and 2015BAK35B01); the Beijing Municipal Science and Technology Project (D161100003016002); the “111” Project under Grant B08043. Correspondent author: YU Zhangguo, yuzg@bit.edu.cn Submitted/Accepted/Revised dates: 2016-07-04/2017-02-16/2017-03-31