Helping robots find themselves–A novel Wi-Fi-assisted localization framework
This new approach solves the performance issues of existing methods and greatly reduces computational costs.
Daejeon, December 11, 2019
A pair of researchers from ETRI and KAIST have devised a new framework that allows mobile robots to accurately determine their position in large spaces by combining Wi-Fi signal information and sensor data. This novel approach, which was published in ETRI Journal, greatly reduces the cost of localization schemes and the associated computational complexity, which is key for widening the application areas of mobile robots, such as surveillance, patient guidance in hospitals, delivery, and other industry-related tasks.
In the last few years, the scope for applications involving service robots has greatly expanded. It won’t be long before robots perform all sorts of delivery, surveillance, and guidance tasks in large buildings such as hospitals, malls, and factories. However, for this to possible, robots need to be able to tell their exact location within the building, which is not a straightforward process. “The navigation system of mobile robots consists of various technologies, such as localization, mapping, recognition, perception, path planning, and motion control,” explainsProfessor Hyun Myung, head of the KAIST Robotics Program.
Although there are methods for the self-localization of mobile robots in the industry field, these approaches generally rely on static physical markers placed in strategic points in the building that robots can perceive. These approaches are not only costly and sometimes complicated to implement, but are also unsuitable in very dynamic workplaces involving many robots, machinery, and people and where things may change location very rapidly.
Other available methods based on particle filters employ statistics and only require a static map of the building. In these methods, the actual position of the robot is derived from a large number of candidate positions that are randomly generated at first. Good candidate points are filtered according to the likelihood of them being the actual position of the robot; this is done using data from the previous known position of the robot and by combining it with sensor data from, for example, laser sensors and an odometer installed on the robot. The problem with such approaches is twofold. First, the initialization step is prone to failing because data from the previous position is not known. Secondly, to compensate for the previous issue, a huge number of candidate positions are generated, which tremendously increases the computing power required.
To address these limitations, the researchers took advantage of the fact that Wi-Fi networks are used virtually everywhere in large buildings and are already part of most robotics implementations. In Wi-Fi communications, there is a parameter called “received signal strength indicator (RSSI)” that is part of the core Wi-Fi standard protocol. This parameter indicates the strength of the Wi-Fi signal received from the router (think about the “signal bars” in smartphones or laptops when connecting to Wi-Fi or mobile networks). Because signal strength is coarsely related to the distance to the router, this information can be used to make an initial estimation of the location of the robot according to the known position of the Wi-Fi routers in the building, thus greatly reducing the area where the position candidates have to be generated.
The performance of the proposed approach was confirmed experimentally using a mobile robot. “The proposed method can solve the problems of conventional particle filter algorithms, both local minimum and heavy computation, by exploiting the ubiquitousness of pervasive networks, such as Wi-Fi or mobile networks,” concludes Dr. Yu-Cheol Lee, senior researcher from ETRI. The proposed low-cost localization method will hopefully help in the gradual adoption of autonomous robots to improve our daily lives.
Reference
Title of original paper: Hierarchical Sampling Optimization of Particle Filter for Global Robot Localization in Pervasive Network Environment
DOI: https://doi.org/10.4218/etrij.2018-0550
Name of authors: Yu-Cheol Lee1, Hyun Myung2
Affiliations:
1SW and Contents Research Laboratory, ETRI, Daejeon, Rep. of Korea
2School of Electrical Engineering, KAIST, Daejeon, Rep. of Korea
About Dr. Yu-Cheol Lee
Yu-Cheol Lee obtained his BS degree from Yonsei University in 2004, his MS degree from POSTECH in 2006, and his PhD degree from KAIST in 2020. Since 2006, he has been working as a senior researcher with the robot research group at ETRI. He has participated in numerous large-scale research projects, performing leading roles as a researcher and manager. Currently, his research interests include the localization and map building for intelligent vehicles and the navigation technology for pedestrians in indoor and outdoor environments.
About Professor Hyun Myung
Hyun Myung obtained his bachelor’s, master’s, and PhD degrees from the Korea Advanced Institute of Science and Technology (KAIST) in 1992, 1994, and 1998, respectively. After being a researcher in the Electronics and Telecommunications Research Institute (ETRI) and in Samsung Advanced Institute of Technology, he became a Professor with the Department of Civil and Environmental Engineering at KAIST. He is also the head of the KAIST Robotics Program, where he researches on topics such as simultaneous localization and mapping, robot navigation, artificial intelligence, swarm robots, and structural health monitoring using robots.
Media contact:
Yu-Cheol Lee: yclee@etri.re.kr, http://sites.google.com/site/yucheollee
Hyun Myung: hmyung@kaist.ac.kr, http://urobot.kaist.ac.kr/
