On December 22nd, 2021, Huawei Smart Car Solution Ecological Forum was grandly held online, with the theme of "Enjoying Ecology and Creating the Future Together".

In the afternoon’s "Dialogue with Che Yun" topic, three experts from Huawei Smart Che Yun shared the "Troika" that drives the development of intelligent networked vehicles in Huawei’s vehicle networking field-OTA, VHR, and three-power cloud services.

Now Huawei has officially released the full text of the speech. Let’s take a look at it here.

With the rapid development of new energy vehicles, under the trend of intelligence and networking, software-defined vehicles have become the industry consensus, and OTA has become the core power of software-defined vehicles. Huawei released the OTA 3.0 vehicle-level upgrade scheme, which can provide a vehicle-level upgrade experience with zero bricks and good experience.

According to reports, the development of OTA will go through three stages. From the market point of view, most of the new forces ICV have taken the lead in entering the OTA 3.0 stage, and maintained a high iterative speed in ADAS/ADS, vehicle control domain, cockpit domain and power domain.

Three core issues that OTA3.0 needs to pay attention to: vehicle version management and quality care, OTA security and reliability, and user experience.

Three core issues that OTA3.0 needs to pay attention to:

Vehicle version management is a systematic problem, how to ensure the compatibility and consistency of vehicle parts versions. At present, the number of lines of vehicle software code reaches 100 million, and the complexity of multi-component collaboration is high. After-sale vehicles have a variety of parts version combinations, which lead to thousands of vehicles. How to ensure the compatibility and consistency of all parts versions of the whole vehicle, how to ensure the quality of versions and prevent bugs, especially those that affect driving safety, is a huge project. Taking 20 parts combinations as an example, the number of versions produced by changes caused by functions and problems may reach thousands in the life cycle of the whole vehicle. In the era of vehicle intelligent experience, OTA is becoming more and more popular, and efficient version management has become a key challenge. The OTA platform that can provide effective software version management capabilities has more advantages.

Intelligent driving, vehicle control and power domain upgrade have become the norm, and how to ensure the safety and reliability of the whole vehicle under OTA is extremely critical. OTA upgrade will lead to brick-changing, which will touch the red line of users’ experience, and there are endless cases of complaints and even public opinion caused by brick-changing in the industry.

Huawei systematically analyzed the brick change caused by OTA. The control loopholes, unavailability of core components, lack of remote repair and software bugs in the whole vehicle upgrade process became the biggest incentives for brick change. How to predict all faults that may lead to brick upgrading in the design stage is a systematic problem. It is necessary to analyze and simulate the potential faults in all scenarios and formulate attenuation measures to prevent them.

Huawei mobile phones supported OTA many years ago, and systematically analyzed the failure modes of mobile phone OTA failures, which also ensured billions of reliable upgrades. The experience is worth learning. Similar to mobile phone upgrade, it also faces user experience problems such as traffic consumption, upgrade time and operation interaction in OTA process.

Based on this, Huawei proposes OTA as a core service, which requires three capabilities in the OTA3.0 era:

Open up the whole process from R&D to commercialization (including management actions and upgrade execution). Covers full upgrade scenarios, provides flexible scenario combinations, and can be upgraded in any actual scenario.

Huawei’s OTA solution is a vehicle-level remote upgrade service for Huawei HI automotive solutions. Natural adaptation HI solution upgrade can cover 45+components such as intelligent driving, intelligent cockpit, intelligent electric, intelligent network connection and intelligent vehicle control; At the same time, Huawei’s open car-end upgrade service architecture supports the upgrade of car enterprises’ own parts and three-party parts, and can support the brush writing specifications of car enterprises’ standard parts; At the application level, it supports the upgrade capability of ADAS/ADS applications, algorithms, OS and HarmonyOS native applications.

At the same time, Huawei has built a user upgrade experience that pays equal attention to security, performance and interaction:

Based on many years of experience in ICT software management, this paper provides a One-Track vehicle-level version management system, which realizes the single-line evolution of versions in the OTA era and greatly reduces the management efficiency of complex software combinations.

Based on DFEMA system, which is an excellent practice in safety and reliability design, 2000+faults are identified and analyzed, and fault attenuation measures are taken comprehensively, so as to achieve the goal of safety and reliability of car-gauge OTA and zero brick change.

Relying on Huawei Cloud 2000+ CDN, dynamic scheduling and source station diversion can realize safe and reliable high-speed download capability. Support parallel differential installation and multi-domain parallel upgrade capability, and shorten the vehicle-level upgrade to less than 25 minutes.

Relying on many years of consumer business experience, Huawei provides multi-terminal friendly cooperative interactive operation interfaces for cars and mobile phones, which is convenient for users to upgrade in different scenarios.

In the era of smart cars, various car companies have continuously developed software ecology, constructed and optimized software management processes, and have professional opinions and thoughts on vehicle upgrade reliability, and also accumulated rich experience.

Huawei provides an OTA service platform, and hopes to continue to explore OTA software ecology, management process, reliability, operation and maintenance with various car companies and ecological partners to build a competitive OTA service.

The rapid development of intelligent networked cars has brought two changes to the connection and experience between people and cars;

In the past, when a car broke down or was in an abnormal state, consumers had to go to the maintenance center for inspection. Now, with this large amount of data, the OEM can give early warning and take the initiative to care for customers in real time, even when the fault is still in its infancy. It is also possible to build a plan for some common problems in advance, and make a more accurate ratio model prediction for spare parts, which greatly improves the efficiency of vehicle operation and maintenance.

By mining the value of data, we can provide better services for users and bring more benefits to car companies.

At the same time, with the development of vehicle intelligence, the digital proportion of each body part is constantly improving: the magnitude of vehicle software code is constantly improving, and now the amount of software code is 10 times that of ten years ago, and the amount of software code in the future will be further improved to 300-500 million lines. Every system and component of the vehicle reports various status signals, logs and alarms every day. At present, the data volume of a single vehicle is on the order of 150-200M per day, and it will be further improved with the increase of sampling frequency and the increase of component complexity in the future.

IT House learned that Huawei put forward the concept of VHR with reference to the experience of ICT industry.

VHR stands for Vehicle History Record, which is a data-driven concept based on the whole life cycle. Its purpose is to realize the visibility, maintainability, user care and efficient operation of vehicles on the basis of a large amount of data. These works will greatly enhance the car owners’ experience and user stickiness in the future, and also bring important values and benefits to car companies.

VHR covers many links, such as data collection, data management, data analysis, vehicle state visualization, vehicle fault diagnosis, trend analysis, prediction, improvement, etc. It is a closed-loop system from vehicle arrival to vehicle departure.

Four common application directions:

Huawei said that in order to improve the efficiency and timeliness of detection, it is very important to build a fault detection method based on federated self-learning. At present, many ECU controllers and devices are added to smart cars, and the amount of data is very large. If the cloud-based model is monitored only by the cloud, there will be problems of untimely monitoring in some scenarios, such as poor network, high delay or even no network. Therefore, end-cloud collaboration is a very important choice.

Cloud-side heterogeneous computing resources support model training calculation, fully integrate the capabilities of NPU, GPU and CPU, realize the application scheduling of heterogeneous computing resources integration at the bottom, and improve the efficiency of model training and execution.

Sensitive personal data end-side training (privacy protection), multi-user feature parameter cloud sharing support joint learning (model accuracy), and increase noise data through differential privacy to improve security.

Using high computing power components such as VDC, CDC and MDC on the end side, the anomaly detection model is gathered on the end side for anomaly detection, rapid reasoning, combined with cloud big data model and data supplement provided by suppliers, fault labels are improved, and the accuracy and real-time performance of problem identification are improved.

Huawei also said that the data-based VHR system needs to be built by the whole industry, including car-side capabilities, cloud capabilities, upper-level scenarios and so on.

Driven by national policies and other factors, today’s car electrification is unstoppable. From 2021 to July, the national sales of new energy vehicles exceeded 1.229 million, up by 210.2% year-on-year. New energy vehicles have developed rapidly, but at present, the spontaneous combustion of new energy vehicles has not been effectively controlled. In 2020, a total of 124 accidents were reported. By September 2021, there were 224 car burning accidents exposed by the media, and consumers began to shift from mileage anxiety to safety concerns.

There are many and complicated reasons for thermal runaway accidents of electric vehicles: local heat generation caused by internal short circuit caused by battery cell process defects, electrical abuse, thermal abuse and mechanical abuse, and thermal stability reduction caused by a large amount of lithium precipitation caused by overcharge, low-temperature charging and fast charging, all of which may lead to thermal runaway. Among them, the safety boundary of thermal runaway caused by internal short circuit is dynamic, and whether internal short circuit leads to thermal runaway needs more judging factors, such as short circuit type, short circuit internal resistance, SOC and so on. Therefore, based on electrochemical mechanism and combined with machine learning technology, Huawei has built a more complex relationship model to realize accurate early warning of power battery safety of new energy vehicles.

Based on Huawei VHR data service base and battery pack simulation system, eight applications for power batteries are constructed, including power battery thermal runaway warning, battery fault detection, battery health SOH evaluation, battery remaining life RUL prediction and other applications, to escort new energy vehicles.

In the aspect of battery safety warning, the recall rate of power battery fault detection and thermal runaway warning can reach more than 80%, and the false alarm rate is guaranteed to be in a low range. In the aspect of battery health assessment (SOH), Huawei’s SOH prediction model based on cloud can achieve the life Cycle estimation error of less than 3% and the remaining cycle prediction error of less than 10%, and can achieve 100% traceability of battery problems based on cloud battery life cycle data.

Huawei said that it is necessary to construct massive sample data through the power battery hardware-in-the-loop simulation system and analyze the real vehicle sample data to effectively overcome the problem of small sample data. At present, Huawei has built fault sample data covering ternary lithium batteries and lithium iron phosphate batteries, and built an engineering library of thermal runaway characteristics of 15+ power batteries.

In addition, Huawei also adopted the domain adaptive algorithm to migrate the algorithm model. When the algorithm model is migrated from the simulation environment to the real vehicle, or in different material systems and formulas, the accuracy of the algorithm remains stable.