Development trend of ADAS intelligent hardware and vehicle networking

2017-11-07

Big data, artificial intelligence, cloud computing, Internet of things, mobile Internet surging, each with different attitude to reach and transform the traditional field. Car networking as part of things, with its intelligent vehicle hardware in the traditional automotive end system, pipe system and the system of local or cloud computing background, combined with the development achievements of other technical direction, in the same stride forward singing militant songs.

At present, the domestic ADAS intelligent hardware product development is in full swing, driverless L2, L3 class demonstration products are also common, we combined with its morphological characteristics simple to see the development trend of automotive intelligent hardware and vehicle networking products.

 

一、hardware (CPU):

 

Vehicle intelligent terminal CPU need to meet the requirements of vehicle regulations, Qualcomm, nVidia, Infineon, etc., have launched their own gauge chip, which is the most direct demand for physical layer. On the other hand, for the higher work efficiency, the new chip on the hardware level of the direct adaptation algorithm is also the direction of active efforts, which are demonstrated on the individual characteristics of the algorithm demo. But the mainstream frequency of ARM architecture mainstream CPU is limited in recent years, has been stuck in less than 3GHz, unlike hard disk storage speed, network bandwidth, etc., in the past few years has doubled. The limited frequency, limiting the algorithm or the operating system and the possibility of infinite horizontal expansion, and the high speed of the network is in turn brought to the terminal processing background online real-time processing may not be limited to end system, offline management system operation, can save the offline CPU pressure calculation.

 

二、Operating system:

 

The requirement of real-time embedded operating system is imperative. Reliability, stability of processing capability and low power consumption are essential. The complex (high speed transmission and the complexity of mechanical vibration, EMI, high temperature, real-time condition etc.) and intelligent hardware vehicle testing to harsh automotive electronic products from the original simple processing logic to the complex algorithm of the system platform, through the ordeal. As a result, the real-time output requires that the delay of the system must be small enough to be millisecond. ECU, the original control unit of automotive electronic equipment, is not a single chip embedded system, but still can not meet the growing demand for operating system in a large number of tasks, but with the operating system, but the corresponding requirements for the delay.

 

三、Terminal algorithm chapter:

 

Vision based or radar based ADAS algorithm, or the implementation of unmanned aerial vehicles, higher Level fusion algorithm and control strategy, this is the core of intelligent terminal. Numerous model training and sample accumulation are only more efficient and useful algorithms library. Depth learning (deeplearning) goes deep into all fields of artificial intelligence, and brings new development ideas for vehicle algorithm. And even think that vision based ADAS can go beyond Mobileye is here - because the technology accumulation of hardware, software, and data sources has fallen behind, and it may be a new direction to dig with a new tool.

The calculated from the local and online, local operation is good algorithm in real time, does not need the network environment, can be free, independent work in various scenes, although is not fixed, but not all things may lead to the uncertainty in the process of the work and can not evaluate. The advantage of online is that powerful background can be used as background. In addition to vertical depth calculation and even horizontal comparison, parallel processing of large amounts of data brings statistical opinions, which is not available for single equipment of local operation, and can move from system to ensemble.

 

四、Network articles:

 

As a basic demand for travel, whether a vehicle starts to go out or not depends on the driving state of other vehicles A single vehicle should be free and independent. However, when we have to rely on road conditions, traffic environment may have normal driving behavior, we found that V2I, V2V is an objective existence, but we do not know to put them into a network environment for active intervention, collective control. Telematics is simply the function of a bicycle, and V2X is the demand for driverless traffic in complex traffic.

The current intelligent hardware networking can be done for the greater two-way communication of data for the cloud, which is the category of mobile internet. The local demand of the local algorithm may be greater than that of the mobile interconnection, and the local communication on the basis of the mobile Internet may be more time efficient and less stable than the local network. From the perspective of the demand of unmanned, high precision map demand route is the mobile Internet data, but more real-time and accurate; V2X more emphasis on local area network environment, can realize the interconnection of local small environment, than the sensing range of bicycles, but not the whole network. The communication protocols derived from these requirements, such as LTE - V and DSRC, have also been debated in different angles between mobile operators and oems.

 

五、The application of big data:

 

Large data on the cloud computing platform reflects the classification, storage, query and other operation behavior of data parallel processing, biased data processing methods. But for the analysis and mining of data content, it is found that the value of the data itself is the output of the data collected on the cloud system. As an intelligent, Home Furnishing or aviation engines and aircraft engine after collecting data background direct diagnosis, car networking system ADAS, as well as the traditional OBD terminal or TBOX telematics products are mainly collected directly upload real-time condition information, more analysis of road conditions, driving scene feature extraction to upload processing results, increase the sources of information. Dimensions, data and content analysis has also undergone a quantitative change to qualitative change. In addition because the bidirectional transmission network, in addition to the original mobile phone APP through the network based on CAN BUS of the vehicle can be remotely controlled, more likely to be the LBS of various information (weather, traffic condition, etc.) combined with intelligent hardware condition diagnosis to the local, for the whole vehicle condition setting and driving behavior settings or suggestions.

 

六、Relationship with car:

 

When the development of the Internet car and outside, with the local environment, look back only to find the electronic equipment inside the network technology has been in the existing CAN BUS breakthrough, although Ethernet instead of CAN BUS in the car for the existing Tesla devices interconnected ahead in practice, but the overall breakthrough in the industry has not yet arrival. Vehicle intelligent hardware is also forced to join the existing vehicle electronic devices in the presence of another component, to participate in the CAN BUS network, but by the transmission data level is not equal, the transmission efficiency is not high. When the vehicle bus network can realize transmission level Ethernet or similar network, the intelligent terminal can free to a terminal with ESP, EPS, BCU and other traditional control unit of the same level of existence; and in turn, data format and content of the traditional control unit will be able to achieve a new plan in the new bus network so, has completed the deeper analysis, interactive and collaborative control etc..

 

七、Relationship with all things:

 

As a part of IoT, the Internet of vehicles can naturally expect the interconnection of other things outside the vehicle chain. In addition to the use of other information, the basis of its own connection is that intelligent hardware identification of data can meet the needs of other things. Some traditional telematics devices, such as GPS, can contribute to the judgment of traffic conditions, and the role of real-time data information of vehicle location in road traffic light settings is often mentioned. From the video, the most direct is the original DVR video for the accident and other effects of reduction.

It's actually based on visual ADAS analysis, and you can dig more. For example, in the traffic accident of the vehicle's speed judgment, in addition to a speed measuring device, on the road of their vehicle accident several electronic devices in ECU will be recorded, if the two were related to video analysis can be considered independent sources of information based on. The implementation can be the video of the vehicle around the accident, or the video based on the networking data to find the vehicles on the local road at that time. Another example may be a better representation of the combination of vehicle born in motion characteristics and the results of intelligent hardware analysis. The traditional map data acquisition part of the way is to use the real vehicle plus a high cost professional data acquisition device in unfamiliar sections or updated sections of field acquisition, but this method is difficult to fully cover the more remote but the data needs of the road section. Based on the GPS data information when the vehicle passes through this section, it is just the route information, but based on the ADAS video recognition, it can be the identification result of traffic information such as lane, street sign, traffic light and so on. On the basis of networking, low cost intelligent terminals, combined with selective local vehicles, can be used as the acquisition source to collect the basic map data more efficiently.

 

Epilogue:

 

As the most promising new carrier of network services after mobile phones, vehicles bear the important mission of developing new forms and contents of services. The mobile Internet, the Internet of things gives its network characteristics, artificial intelligence gives it the possibility of unmanned driving, two directions have brought its own big data generation and its demand for big data. At present, the intelligent hardware and its associated vehicle network can also be considered in the initial stage, the completion of the traditional transformation, as well as the new functions, new content increased car will be a rich and colorful existence, we look forward to in practice.