The advent of the era of intelligent vehicles has also brought about the outbreak of various types of on-board sensors, and lidar, millimeter-wave radar and cameras are recognized as the core sensors of the three major unmanned vehicles. In order to achieve higher-order automatic driving, lidar is undoubtedly the most important incremental accessory, Google, Audi, Ford and Baidu and other companies have developed unmanned vehicles have been equipped with lidar, Weima, Weilai, Zhiji, Xiaopeng and Nezha and other new car-making forces have also released lidar-based products and development roadmaps this year. At present, lidar is gradually transitioning to commercialization and large-scale commercialization, and is paving the way for the final mass production of unmanned vehicles.
According to CICC' forecast, The size of China's lidar market will climb rapidly from 5 billion yuan to nearly 40 billion yuan in the next five years. By 2025, the global lidar market size is expected to rise from the current around $15 billion to the 100 billion level. The reason why lidar will fully explode is fundamentally because vehicle-mounted lidar has better performance than cameras, millimeter-wave radar, etc. in terms of ranging accuracy and anti-jamming. It is worth mentioning that the current vehicle-mounted lidar gradually enters the growth period from the introduction period, the uncertainty begins to increase, and the choice of future technical routes, the development of mass production customers, and financing capabilities will affect the competitive pattern of lidar.
Companies in the United States, Germany, Israel and other countries have started early in the development of lidar and have mature technology. Due to the high technical barriers, there are not many "players" in the domestic laser radar track that mainly pursue research and development for driverless cars.
But laser quantum ghost imaging radar, Chinese technology is at the forefront of the world.
Quantum imaging has a history of more than 20 years since it was first realized in 1995. At present, quantum imaging has completely abandoned the entangled light source, stepped out of the demanding laboratory environment, and formed a multidisciplinary intersection with various fields, including compression perception, and was further introduced into the field of laser detection. Laser quantum ghost imaging radar can achieve field segmentation and target recognition without relying on GPU or AI chips in the process of optical detection under the framework of traditional neural networks, which can significantly improve the imaging and recognition effect of targets under bad weather conditions, and its low-dose high-sensitivity detection performance can provide a strong guarantee for equipment and human eye safety in future unmanned traffic situations. On this basis, the quantum ghost imaging can be combined with the addressable vcsel array to achieve the true solid state and mass production of the laser ghost imaging radar.
Recently, 36Kr came into contact with a domestic innovative enterprise in the field of laser quantum ghost imaging radar, Shenzhen Yiwei Ruiguang Technology Co., Ltd. (hereinafter referred to as "Yiwei Ruiguang").
Researcher Han Shensheng of the Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, is an internationally recognized authoritative expert in the field of quantum ghost imaging radar. Dr. Wang Xiaopeng founded Yiwei Ruiguang based on the previous academic cooperation with researcher Han Shensheng in the field of quantum imaging, and absorbed a group of industry "veterans" including the marketing and sales team of international first-line lidar companies.
Wang Xiaopeng's team believes that the technology route of solid-state laser quantum ghost imaging radar, which has China's independent intellectual property rights, will become a key breakthrough. Solid-state laser quantum ghost imaging radar will bring real solid-state, mass-produced and safer products to the market, greatly alleviate the tension of existing unmanned vehicle computing power, ensure human eye safety and equipment safety in unmanned traffic flow scenarios, and provide low-priced, highly reliable, all-weather, all-weather imaging lidar for the unmanned logistics industry.
Will the technical barriers and practical challenges of lidar be broken by Chinese? With many questions about this cutting-edge technology field, 36Kr and Wang Xiaopeng, founder of Yiwei Ruiguang, conducted an in-depth discussion.
Put a can in front of the car, how can the unmanned car judge what it is? For a vehicle equipped with high-level assistance and autonomous driving capabilities, it is first scanned and imaged by the on-board camera and lidar, and then the obtained images and point clouds are fused. The fused data is further fed into the car GPU or AI chip, where the pre-trained neural network determines whether it is a cup, a mobile phone, or a can.
Because the road conditions faced by unmanned vehicles are often very complex, the amount of data they need to process is very large. According to the core technical personnel of a domestic line L4 automatic driving company, at present, more than 60% of the computing power of the on-board computing platform is consumed in the processing of the lidar point cloud by the perception algorithm to cope with complex traffic flow and road conditions. Therefore, while further optimizing the consumption of resources and computing power by related algorithms, major companies have given very high attention and expectations to higher performance GPU or AI chips. Nvidia, Tesla and Intel have introduced new and stronger in-vehicle chips every year, and domestic latecomers such as Horizon and Cambrian are not willing to show weakness, and the computing power has been continuously improved, while pushing up the valuation.
Solid-state laser quantum ghost imaging radar can solve the high requirements of unmanned vehicles for the computing power of on-board chips
The root of the high demand for the computing power of the on-board chip of unmanned vehicles is that the real-time point cloud data of lidar is too large, so its transmission and processing are limited by hardware performance. Different from the simple solution of some existing lidar companies (lidar does not output point clouds and outputs simple identification results), the solid-state laser quantum ghost imaging radar of Yiwei Ruiguang transports the neural network originally running on the GPU or AI chip to the process of optical detection, while detecting while segmenting the field of view and identifying the target, thereby eliminating the hardware transmission of the point cloud and the processing on the vehicle chip, which greatly alleviates the computing power requirements of the unmanned vehicle on the vehicle chip.
The technical basis on which the Evi ruiguang solid-state laser quantum ghost imaging radar relies is the work that Dr. Xiaopeng Wang and researcher Han Shensheng collaborated on between 2016 and 2017: "At that time, Mr. Han and I were testing the performance of error control coding in the field of communication to reconstruct the target in the imaging scene. However, in the follow-up experiments, we realized that in fact, this idea with interdisciplinary characteristics is essentially to move the neural network to the process of optical detection," Wang Xiaopeng further explained: "Unlike the existing line scan mode lidar products on the market, our quantum ghost imaging radar can not do any spatial information solution for cars, people or obstacles, and perform light-end neural network operations on cars, people, road surfaces or obstacles by detecting the combination of node information in the light field. Directly obtain the results of segmentation and identification of these targets. This is one of the most characteristic properties of quantum ghost imaging radar. ”
According to the type of scanning, the common lidar products on the market are mechanical rotation scanning, mems scanning, rotational reflection and refraction prism scanning, optical phased array opa scanning and flash scanning. According to the principle of ranging, the common lidar products on the market are pulse ranging type tof class, frequency modulation continuous wave fmcw class, triangle ranging class and so on. In essence, solid-state lidar is a lidar that has no moving parts at all in the scanning mechanism, and generally does not refer to a specific ranging principle.
At present, the more popular lidar in the market is based on the mems principle, the OPA principle and the flash mode. From the current situation, mems and lidar based on the principle of rotary folding reflective devices still do not fully meet the standards of vehicle-level mass production, which is also related to the fact that these radars have more or less large or small size moving parts. OPA and flash mode lidar are more likely to become vehicle-grade sensors for unmanned vehicles, but the former always has the problem of low system efficiency, and the latter also faces the problem of limited supply of core devices. High-performance multi-channel SPAD arrays are also in the hands of monopoly giants like Sony, ON Semiconductor and Hamamatsu.
Addressable vcsel array light source
In such a context, another outstanding feature of Evi ray Light's solid-state laser quantum ghost imaging radar is that, unlike existing lidar products, the system structure of quantum ghost imaging radar is based on a programmable array light source and a single-pixel receiver. Therefore, for the line sweep scheme, the introduction of the programmable array light source solves the problem of obtaining point cloud spatial information while ensuring the solid state, while for the flash scheme, the use of single-pixel reception removes the dependence on the high-performance multi-channel area array detector of foreign manufacturers. As a typical representative of programmable area array light sources, the R&D team of Everbright has rich design experience in addressable vcsel array light sources.
"With the blessing of mature carrier-grade processing and packaging and testing capabilities in China," Wang Xiaopeng added, "our solid-state laser quantum ghost imaging radar will bring real solid-state, mass-produced, reliable and low-cost products to the market." ”
With the gradual rollout of autonomous driving technology, more and more unmanned vehicles equipped with lidar have appeared in the public eye. In the foreseeable 2-3 years, the phenomenon of unmanned traffic will appear on the road. Assuming that an unmanned vehicle is equipped with 3 lidars, then the extremely dense lidar point cloud will appear along with the unmanned traffic flow, at this time not only the safety of the human eye, traffic cameras or public security audio and video acquisition equipment will face a great security threat, and the crosstalk between lidar will become very significant. It is no exaggeration to say that the scene at that time will probably be that different unmanned vehicles will become unreliable due to the mutual interference of lidar, the existing traffic and public security equipment on the roadside will become unavailable due to frequent interference, and even the vision of pedestrians and human drivers will be threatened.
"Unlike existing lidar solutions, quantum ghost imaging radars inherently have the ability to detect under low-illumination conditions," Wang Xiaopeng explains, "This point was fully demonstrated theoretically and experimentally by our team members in the laboratory of the institute in the early days, and the relevant results were also reported by science magazine." Our actual measurement results show that under the quantum ghost imaging radar system, we can use one thousandth of a radiation dose to complete the same quality of target image acquisition. In other words, the installation of laser quantum ghost imaging radar can significantly reduce the adverse effects of high-density point clouds generated by unmanned traffic on pedestrians, drivers and roadside audio and video acquisition equipment, and truly achieve safety. ”
For laser quantum ghost imaging radar, due to the particularity of its space-time correlation detection method based on the light field, there is theoretically no other signal with correlation at the same time in addition to the irradiated light emitted by itself. Therefore, laser quantum ghost imaging radar can also cope with the problem of lidar crosstalk that accompanies unmanned traffic.
According to Wang Xiaopeng, in the actual application process, the performance of lidar products on the market will be significantly affected by the working environment. In the good weather conditions can play their own performance relatively normally, but in the more severe cases, such as typhoons, heavy rain, smog, heavy snow or sand, etc., the existing lidar products measuring distance significantly reduced, the emergence of many false targets that are difficult to identify, in some extreme cases there will be complete blindness, resulting in a sudden drop in the performance of the unmanned vehicle perception system, or even complete failure, vehicle emergency parking.
All-weather detection capability of solid-state laser quantum ghost imaging radar
For many operating vehicles or private cars, it may not be necessary to take risks in bad weather conditions, but for special application scenarios such as logistics, emergency response, disaster relief and defense, all-weather performance is a basic requirement. The solid-state laser quantum ghost imaging radar of Everbright Light can give full play to the characteristics of high-dimensional detection of quantum imaging, and has good resistance to light field attenuation, scattering and turbulence under bad weather conditions. "This feature is that the core members of our team have also conducted field tests on short, medium and nearly three scales in the real environment under the research and development framework of the national twelfth five-year plan, and the effect is obvious." Wang Xiaopeng said.
In terms of market strategy, Wang Xiaopeng pays attention to reality: "Although we are based on quantum technology, when considering the market, we actually focus on logistics. The logistics mentioned here include trunk freight, transshipment and distribution vehicles, and unmanned warehousing. The common feature of these segments is that the market is large, highly concentrated, and easy to promote and popularize on a large scale. At the same time, the requirements of the logistics market are simple and crude, cheap, reliable, mass production and all-weather. ”
"In fact, I am very optimistic about the technology of unmanned vehicles, and I am convinced that it must be one of the key technologies that can change the shape of human society in the next ten or even twenty years," Wang Xiaopeng continued: "For today's China, especially after experiencing various natural disasters such as the new crown pneumonia epidemic and the heavy rains in Zhengzhou, the efficient storage, transshipment, concentration and distribution of materials shown by China's logistics network have marked that this industry has become an important support for the national economy and people's livelihood." It is of great strategic significance to build an efficient, reliable, robust, all-weather, all-weather unmanned logistics network. Therefore, in such a context, the deep combination of solid-state laser quantum ghost imaging radar with all-weather detection capabilities and unmanned logistics is inevitable. ”
According to the "Terminal Unmanned Delivery Track Research Report", at present, the most important scenarios of terminal unmanned delivery are express delivery and instant delivery, the latter including takeaway, fresh food, supermarkets, retail, etc., no matter which scene, it contains huge market value. The report believes that in 2021, the scale of China's terminal distribution market will exceed 300 billion yuan, the unmanned distribution business model has been initially formed, the technology has been able to support the demand of some scenes, and the industry has entered the eve of batch commercial applications. The report predicts that the world's first unmanned fleet of more than 1,000 units will be born in the unmanned delivery track this year. In the next 3 years, the cost of unmanned delivery vehicles will gradually drop to less than 100,000 yuan, and the unmanned delivery track will take the lead in ushering in an outbreak. This gives Yiwei Ruiguang's "solid-state laser quantum ghost imaging radar" product an excellent entry path, "we hope that the market share is more than 10%," Wang Xiaopeng said: "But in business logic, we do not rule out working with partners to enter other areas with advantages." ”
In terms of commercialization, Yiwei Ruiguang plans to launch its first generation of short- and medium-range solid-state laser quantum ghost imaging radar for major unmanned logistics and distribution enterprises at home and abroad by the end of 2022. On the basis of prioritizing cost, reliability and mass production capabilities, we will gradually provide the market with upgraded products with 24/7 detection capabilities and optical end neural network computing capabilities. At the same time, with the full rollout of L3 autonomous driving and the commercialization of L4 level robotaxi, Yiwei Ruiguang plans to further expand its product range to the passenger car market around 2024.
The core technology of Yiwei Ruiguang was inspired by the collaborative research of Dr. Wang Xiaopeng and Han Shensheng's team, the first echelon of quantum radar of the Institute of Optical and Mechanical Research of the Chinese Academy of Sciences, during his study at the University of Sydney in Australia. There are many head players in the domestic lidar track, such as Shanghai's Hesai, Shenzhen's Sagitar, Beijing's Beike Tianhua, Huawei and DJI also announced their entry last year, but the solid-state laser quantum ghost imaging radar technology route is currently only one yiwei Ruiguang.
The marketing and sales teams of Yiwei Ruiguang are from international first-line lidar companies, who have been in the industry for many years and are familiar with customer groups such as OEM and tier1. At the same time, Yiwei Ruiguang also absorbed a product development team with rich engineering experience, with a number of doctors and high-end talents returnees, with an average experience of more than five years, with implementable project development experience.
It is reported that Yiwei Ruiguang has recently officially launched an angel round of financing, and the funds are planned to be used for prototype production, technology investment in Wuhan Optics Valley and Shanghai R&D base, national market and customer development, etc., to quickly promote the landing of mass production and industry application scenarios. The vision of Yiwei Ruiguang is to become a cutting-edge technology enterprise that accelerates the industrialization of large-scale quantum imaging technology and the landing of application scenarios, completely solves the above pain points that have existed for a long time in the industry, and brings technological upgrading to the lidar industry.
According to the prediction of the globally renowned economic consulting agency ihs Global Perspective Automobile Division, the shipment of autonomous vehicles is expected to reach 54 million units in 2025, and the loading volume of lidar is expected to reach 15 million units, or about 100 billion yuan. During this period, Evi ruiguang hopes to quickly cover the market and become the forerunner of solid-state mass production products in the lidar industry.