编程知识 cdmana.com

Accurate! Wuhan University and Gaode team won the international top competition "mobile phone positioning in vehicle environment"

How to have a more accurate mobile navigation experience ? Now people are more and more used to using mobile phones to navigate the route when they go out . The experience is faster and more accurate than in previous years , In addition to the improvement of mobile phone hardware performance , A whole set of technical solutions played a key role , And positioning is the most basic technology . This is also an important area of research by the Golder technical team .
 
In the just concluded International Summit on Positioning IPIN2020 In the game , Wuhan University and the positioning R & D Department of Gaode intelligent technology center jointly organized a team to participate in Track 6 Group competition . The purpose of the competition is Mining is based on smart phone built-in sensors ( Include GNSS, gyroscope , Accelerometers and magnetometers ) The potential of vehicle integrated navigation and positioning scheme , The test path includes outdoor open scenes 、 Partial occlusion scene and indoor satellite signal occlusion scene . Final , Wuhan University and Gaode United won the championship by an absolute advantage .
 
International indoor positioning and Navigation Conference IPIN(Indoor Positioning and Indoor Navigation) From 2010 year , It is the largest international academic conference in the field of indoor positioning , The competition held at the same time of the conference is one of the three major indoor positioning competitions in the world , The other two events are hosted by the National Institute of standards and technology (NIST) And Microsoft ,IPIN The conference is one of the top positioning and navigation conferences in Google's academic rankings .
(https://scholar.google.com.au/citations?view_op=top_venues&hl=en&vq=eng_radarpositioningnavigation)
 
Combine with real scenes to enhance the difficulty Test the calculation ability of algorithm inertial navigation
 
With the popularity of smart phones , There are a large number of users using mobile phones for positioning and navigation , There are many challenges in using mobile location , Mainly depends on GPS location , because GPS Signals are vulnerable to NLOS 、 Multipath interference , There's a lot of lost stars 、 Drift and so on , Especially indoor scenes ( Tunnel 、 Parking lot, etc ).
 
At present, sensors and GPS The fusion positioning method maintains the stability and continuity of positioning . The components of mobile phone are limited by cost , Often the accuracy is limited , It is difficult to meet the minimum requirements of vehicle navigation and positioning . therefore , In this year's IPIN In the game , It is the first time to join the mobile positioning project in the vehicle environment .
 
After the game (off-site) The way , The organizer provides mobile phone data on the way of driving navigation ( Include GPS、 The acceleration 、 gyroscope 、 Magnetometer 、 Air pressure, etc ), Participants restore the actual position of the vehicle according to the mobile phone data .
 
Make an appointment for the whole audience 19 km , The mobile phone is fixed on the car , front 4.5 Km is used for team algorithm initialization , after 14.35 Km is used to verify the algorithm , There will be periods of varying length in the validation phase GPS Signal loss ( in total 23 minute ), as well as 11 Minutes indoor area . In the initialization phase , It is mainly used for sensor correction 、 State initialization and installation angle estimation , The initialization trajectory includes multiple turns 、 Stop and wait . In the validation phase, we take Artificial shielding GPS The signal ( simulation GPS Unable to locate ) and Driving indoors ( Restore the real GPS Unable to locate the situation ).
 
Single time in the game GPS The signal is interrupted for a long time , In the validation phase , share 19 Time GPS Signal interruption , among , The longest interruption time is 400 More than a second ,GPS When you can't locate for a long time , It is a test of the algorithm's ability to calculate inertial navigation , Including the accuracy of sensor bias estimation 、 The error size of state estimation, etc . stay GPS During signal interruption , No other offers absolute position 、 Speed signal (WIFI、 The base station 、 bluetooth 、 Odometer, etc ). therefore , Once there is a big deviation in the algorithm , Will continue to affect the positioning after . The last two long periods of the trajectory GPS Interruptions pose the biggest challenge , One of them is to drive freely in the open airport , Can't use road network matching technology . It is very difficult to restore the real position by using sensor inertial navigation alone .
 
Last , With the help of big data mining and artificial neural network , Using relatively new VDR The core technology , Combined with the special movement mode of vehicle, the model and equation are designed , Reduce the calibration deviation , It effectively avoids the defect of insufficient hardware performance , Make full use of the advantages of data post-processing to further improve the accuracy of track position, and won the championship .
 
Joint efforts to tackle technical difficulties Enhance navigation user experience
 
DR(Dead Reckoning) It is an important technology in navigation and positioning , The technology integrates inertial measurement units (IMU)、 Multiple information such as speed , Realize in GNSS Continuous position estimation in case of signal loss . GORD is in the car DR The accumulation of technology in the field started earlier , Leading the industry , At present, it has been applied to hundreds of models , Guaranteed in the city Canyon 、 Tunnel 、 Continuous and accurate navigation in complex scenes such as parking lots .
 
In recent years , Gaode positioning research and development team has been committed to combining mobile phone sensors and big data technology , Use machine learning and other intelligent technologies to solve the complicated problems of navigation scene . For example, up and down the viaduct 、 Main and auxiliary road identification and other scenes , Use only... In these places GPS It's difficult to make effective location discrimination and accurate navigation , For a long time, it has been the pain point of mobile navigation positioning . Gaode innovates by mining big data of travel , Extract classics VDR The mobile phone posture change information obtained by this technology is regarded as the key feature , combination GPS Information 、 satellite SNR Information, etc , Representation learning for local road sections , Effectively solve the above problems . Relevant functions have been launched on Gaode map .
 
Compared with the car engine DR, mobile phone DR Even greater challenges . First , On the way to car navigation , Mobile phone posture 、 The position is often not fixed , For determining whether the mobile phone is fixed on the bracket 、 It is very important to estimate the posture of mobile phones . secondly , There's no odometer on the phone , stay GPS When unable to locate , It is difficult to accurately estimate the actual speed of the vehicle , Lead to VDR The calculated position is easy to diverge . Last , There are many kinds of smart phones , There's a huge difference in cell phone sensors , How to ensure the positioning accuracy of different mobile phones 、 Reliability is also an important issue to consider .
In order to improve the positioning accuracy of mobile phone in complex environment , So as to improve the travel experience of Gaode users . Gao de and Wuhan University are jointly working on the key technology of mobile phone dead reckoning (VDR) And has made a lot of innovative breakthroughs , I believe there will be more good news in the near future .
 
 

版权声明
本文为[amap_ tech]所创,转载请带上原文链接,感谢
https://cdmana.com/2020/12/20201224214729014U.html

Scroll to Top