RTK, the anchor of autonomous driving

In the early years of the rapid development of intelligent driving, we could often see towering lidar and white antennas on vehicles.

The white antenna is part of the combined positioning system, the RTK antenna.

Some models are equipped with one pole for high-precision positioning; some models are equipped with two poles, such as Ford, which can achieve positioning + orientation; and even more, such as Cruise, are equipped with more poles, which can achieve positioning + orientation + posture. .

Before 2018, people outside the industry knew little about RTK, and many people did not believe that centimeter-level positioning could be achieved through GNSS.

In fact, 90% of smart driving startups use this technology, and the supplier of this technology is basically the same company: NovAtel in Calgary, Canada.

However, no company will actively mention their use of RTK because the RTK+IMU combination is too expensive.

Novatel's RTK+IMU integrated navigation system, even the lowest-end model, costs around 200,000 yuan.

At the same time, the initiators of this wave of smart driving, the famous Mobileye and Tesla, both started by relying on vision. If RTK is promoted too much, investors will feel that the technology is not high enough, thus lowering valuations.

As time comes to 2022, the industry landscape has undergone fundamental changes.

The rapid rise of domestic smart driving parts startups.

In the field of vision, we have Horizon; in the field of lidar, we have Sagitar Juchuang, Hesai and Tudatong; in the field of RTK, we have Qianxun, Liufen, Hexin, Sinan and Daoyuan.

The cost of RTK has dropped rapidly. Compared with the previous selling price of 200,000 yuan, the cost of parts and components of several thousand yuan has made OEMs no longer shy about talking about RTK.

Many car companies such as NIO, Ideal, Zhiji, Feifan, and Geely have generously introduced on their product pages that their models are equipped with two high-precision positioning units.

But what exactly is RTK?

Maybe everyone’s first reaction is that this is differential GPS.

Not really.

RTK is the abbreviation of English Real-Time kinematic, which is translated into Chinese as real-time dynamics.

This is quite confusing, what is real-time dynamics? - This statement lacks a subject.

How is RTK technology connected with intelligent driving? What is its position in the intelligent driving system?

Wait for me to come one by one.

01. Raimondi - the father of RTK

Going back to the end of 1972, Lieutenant General Schultz, then commander of the U.S. Space and Missile Systems Center, approached Colonel Bradford Parkinson of the center and wanted to ask him to help promote the research and development of the 621B project. This is the U.S. The beginning of the GPS project.

Interestingly, Parkinson recounted in many of his memoirs his bargaining with Schultz, telling the general that if he could be appointed as project leader, he would do it, otherwise forget it.

So, the general rejected him, and he also rejected the general.

If none of them had compromised, we might have had to wait a few more years for GPS.

Fortunately, when he stepped out of the general's office door, the generous general gave in, stopped him and appointed him as the project leader.

From 1973 to 1993, the United States spent 20 years building a GPS system. After that, Colonel Parkinson was called the father of GPS.

In addition, as early as 1978, Colonel Parkinson planned to apply GPS to driverless tractors, and in 1997 he worked with John Deere Tractor Company to put this concept into practice.

This should be the earliest intelligent driving, 7 years earlier than the first DARPA in 2004. As for why it was the 1990s, that's because RTK didn't really become commercially available until the 1990s.

In the eight years since the first phase of GPS satellite launch, the United States launched 11 Block I experimental satellites.

Professor Charles Counselman of MIT designed the first GPS receiver. The military appointed an equipment factory, Magnavox, to develop the receiver.

Later, TI and Magnavox successively made their own commercial receivers TI-4100 (1981) and WM101 (1985).

But people found that GPS could not meet the previous design indicators (horizontal 3-5m, vertical 10-15m).

The positioning principle of GPS actually measures distance by calculating flight time, so the speed of electromagnetic waves flying in the atmosphere is defaulted to the speed of light.

But the earth's atmosphere is not a vacuum. The complex atmosphere causes refraction and propagation losses of electromagnetic waves.

For example, charged particles in the ionosphere and water vapor in the troposphere are key factors that influence flight time estimates. At the same time, the multipath effect on the ground also leads to a decrease in positioning accuracy.

Another problem is that due to the large size of the receiver, even commercial receivers can only do static surveying and cannot be carried around (kinematic surveying).

This is the origin of the inexplicable name RTK (Real-Time kinematic surveying, high-precision relative position in real time).

However, the real RealTime will have to wait until there are enough Galileo and Beidou satellites in the 21st century.

Before this, because there were many operations, it could only be said to be accurate in real time.

 Professor Raimondi and his doctoral thesis

GPS became available in the 1980s, but RTK technology had to wait until 1992.

The reason is that RTK is waiting for his master, who is Benjamin Remondi.

Raimondi previously worked at NASA and the National Oceanic and Atmospheric Administration, specializing in satellite orbit calculations.

Later, he joined the U.S. National Geodesy and studied for a PhD at the University of Texas at Austin. He wrote the pioneering work in the field of RTK: "Using the global positioning system (GPS) phase observable for relative geodesy: modeling, processing, and results."

This paper uses TI4100 to demonstrate that decimeter-level positioning accuracy can be achieved using carrier differential.

However, this is actually a static measurement. It still takes several minutes to calculate the precise position, and dynamic real-time measurement is not possible.

Moreover, in actual applications, a large amount of in-house post-processing is required.

The RTK algorithm model appeared seven years later in 1991. Mr. Raimondi finally completed OTF (on-the-fly, which uses the Doppler effect and calculates the carrier phase to process dynamic data without static initialization. collected data) implementation of the algorithm.

The OTF algorithm means RTK enters practical applications.

Although there are many researchers at the same time as Raimondi, such as Ron Hatch of NavCom. But because Raimondi made a lot of landmark contributions in the field of RTK, he is recognized as the father of RTK.

A little gossip is that Raimondi and many of his colleagues eventually went to the American company Trimble, which created Trimble's unshakable dominance in the field of GPS measurement in the 1990s and the first decade of the 20th century.

GPSDGPSRTK

02. The story of Tianbao and other pioneers

In 1978, Dr. Charles R. Trimble and two other colleagues left the famous HP company and founded Trimble Company. They began to study GPS technology while at HP, so they started working on receivers with ease.

In 1984, Trimble launched the world's second commercial receiver, the 4000A (the first was TI), and its main R&D force was Javad Ashjaee.

Five years later, in 1989, Trimble acquired the Navigation Systems Department of TAU, which gave them access to DGPS technology.

Please note that differential GPS refers to DGPS, not RTK.

As a commercial company, Tianbao cannot research high technology without financial support. In fact, the father of the funder who really promotes the advancement of RTK technology is the US military.

1988 年，美国陆军工程兵团为了疏浚河道，决定资助开发一台 RTK 原型机。

这个原型机的论证工作交给了加拿大新不伦瑞克大学，同时让天宝负责系统设计。

1991 年，经过了无数天才科学家的努力(主要是雷蒙迪)，RTK 的原理样机终于被造了出来。

随后，工程兵团寻求商业合作，Magnavox 和 Trimble 接下了商业化任务。

1992 年，天宝在工程师 Nick Talbot 的努力下，生产出了第一台商用 RTK 接收机 4000SSE。

Nick 在澳大利亚攻读博士期间就仔细研读了雷蒙迪教授的博士论文, 并和他有诸多交流。

Magnavox的RonHatch博士也搞出来一个样机。

但是很不幸，Magnavox 没有坚持到最后，早在 1994 年就卖给了徕卡，后来徕卡成了 RTK 市场的另一大玩家。

1992 年，请记住这个年份，因为国产的 RTK 接收机是 2012 年，这可是二十年以后的事情。

如果雷蒙迪是RTK算法之父，那么贾瓦德算的上 RTK 接收机之父。他首先在天宝搞出了 T4000S，然后 1989 年自己创业成立 Ashtech，这个公司几经易手，2011 年最后卖回给了天宝。

1998 年，他又成立了 JAVAD Positioning Systems，然后卖给了 Topcon。

然后，2007 年，他又创立了 JAVAD GNSS 公司。

一个人带起了业内好几家重量级的玩家。

 1992 年，天宝将 RTK 商用化

GPS 在美国的迅速发展也刺激了北方邻国加拿大的 GPS 产业。

1980 年，在加拿大的卡尔加里，一名叫杰拉德 (Gérard Lachapelle) 的工程师与朋友一同建立了 Nortech (Surveys) Canada，后来又建立了 Norstar Instruments, 后来这家公司就改名成了大名鼎鼎的 NovAtel。

几年后的 1988 年，杰拉德教授加入卡尔加里大学成为了一名教授，并一手创立了卡尔加里定位导航研究小组（Position, Location And Navigation (PLAN) Group）。

1989 年，这个小组开始研究差分 GPS 技术。

值得一提的是，杰拉德教授夫人 M. Elizabeth Cannon 教授与他在同一个实验室工作，并在 2010~2018 年成为了卡尔加里大学的校长，足见 GPS 方面的研究对于卡尔加里大学学术声誉的影响。

1992 年教授夫妇和 NovAtel 公司的 Patrick C. Fenton 联合发表论文，说明他们一家完成了 RTK 原型样机的开发，仅仅比天宝的进度晚了一年。

有点遗憾的是，诺瓦泰于 2008 年以 3.9 亿美元的价格卖给了海克斯康。

而收购了 Magnavox 的徕卡地理信息部门也早在 2005 年就以 14 亿瑞士法郎的价格卖给了海克斯康。

GPS 的两大先驱公司就这样投入了不差钱的海克斯康门下。

不过我搜索了一下，就算在瑞典，海克斯康也不算是一家顶级的公司，年销售收入在 45 亿欧元左右，居然排不进中国企业 500 强，但是人家含税净利率是真心高，达 30%。

 诺瓦泰关于 RTK 的研究

 RTK 接收机供应商发展图

03、CORS 站

在美军工程兵团推进 RTK 接收机研发的同时，美国海洋与大气管理局下属的国家地理测绘局已经开始建立国家空间参考系统（NSRS）。

1987 年，美国地理测绘局在田纳西州建立了一第一个参考站，名字叫做国际合作 GPS 网络 (CIGNET)。

他们使用了高精度双频接收机，7X24 小时连续记录卫星信号。

到了 1994 年，鉴于很多机构都在做相同的事情, 例如美国海岸警卫队，美军工程兵团，联邦航空管理局，喷气推进实验室（钱老创立），有专家就呼吁建立一个统一的网络。

最终，在 CIGNET 的基础上，建成了美国的 CORS( Continuously Operating Reference Stations , 连续运行参考站系统) 系统。

目前，美国的 CORS 系统有 3050 个站点，广泛的分布于美国、加拿大、太平洋还有波斯湾。

日本的 CORS 网络起步也很早，1994 年，日本就建成了第一个基准站。

截止 2007 年，日本 GEONET 有 1308 个站分布在全日本各地。而欧洲建站则相对较稀疏，全欧洲只有 398 个固定站。

当然，随着技术的发展，我们又有了 PPP-RTK 技术，可以单机计算时钟误差、轨道误差、相位偏差，并通过接收卫星播发改正数来计算电离层和对流层延迟。

于是就有了 Omistar、RTX 、Starfire 、TerraStar 这样的播发系统, 通过播发卫星就可以覆盖全球，日本的天顶卫星更是自带 PPP-RTK 支持。

 RTK vs PPP

PPP-RTK无需建设高密度的地面参考站也可以实现实时分米级别的定位。

但是相对于 RTK 也有着接收机硬件要求高，改正数数据量大，初始化时间长，受到天气及太阳活动影响大，定位精度容易降级等一系列的问题，尤其是建立实时动态的高精度本地电离层模型是其面临的一大挑战。

有专家认为，在 RTK 基准站密集的地区，使用 PPP-RTK 技术意义不大。