JD, Neusoft and Intel Jointly Building Intelligent and Connected Vehicle Cloud for HaiMa(former Hainan Mazda)


Intelligent interconnection is currently regarded as an important development direction of the auto industry in China, and in-vehicle intelligent applications are set to be at the cutting edge of R&D for new models. In order to enhance product competitiveness, Haima has built an intelligent and connected vehicle cloud to support the production of the new-generation vehicles.


Development of the Haima Cloud Platform for Intelligent and Connected Vehicles is undertaken through close collaboration among JD, Neusoft and Intel. The system provides services like roadside assistance, maintenance and recall reminder, vehicle status and positioning, vehicle maintenance and repair, transmission of information, electronic manuals, etc., and supports weather, music, navigation, and other online applications. Relevant data is collected and analyzed to establish a data foundation for CRM applications. It also supports 3 major infotainment functions - vehicle to home connection, voice shopping, and intelligent maintenance. This new cloud platform for intelligent and connected vehicles takes driving safety and ease of vehicle-to-home interconnection to a new level.


Vehicle Monitoring and Management System - Everything is under Control

The new cloud platform for intelligent and connected vehicles is based on Neusoft's vehicle monitoring and management system, which is built on the RealSight APM application performance monitoring platform. It boasts industry-leading mass time series, monitoring of the collection, storage and analysis of big data, and visual presentation capabilities. This enhances the performance of the system in actual scenarios as well as the visual experience.

AI-based Analytics and Prediction-Making the system "Smarter"

Analytics Zoo is a unified platform for big data analytics and artificial intelligence open sourced by Intel. It integrates Tensorflow, Keras, PyTorch, Apache Spark, Apache Flink, Ray programs, etc., in a integrated pipeline, and can be transparently scaled out from laptop to large-scale clusters for production data processing. The following figure shows the overall architecture of Analytics Zoo.

Analytics Zoo allows users to easily build and deploy end-to-end AI applications. For example, writing TensorFlow or PyTorch codes in Spark programs and performing distributed training and inference; or running Ray applications directly in a big data cluster using RayOnSpark. Using Analytics Zoo, users can make use of the advanced machine learning workflow to automate the development process of large-scaled machine learning applications. For example, users use the automatic distributed Cluster Serving for TensorFlow, PyTorch, Caffe, BigDL, and OpenVINO model inference, or time series furcating via scalable AutoML support. In addition, Analytics Zoo also provides various algorithms and models for building different application scenarios for recommendations, time series analysis, computer vision, and natural language processing.

In a typical machine learning application, users usually have to carry out the appropriate data preprocessing, feature engineering, feature extraction and selection, etc., so that the data set can be effectively used by machine learning applications; and after the data is pre-processed, users must utilize the appropriate models and tune hyper-parameters to maximize the accuracy of machine learning models and algorithms. Obviously, these steps are extremely challenging, making it very difficult for most people to use machine learning technology.

Automatic machine learning (AutoML) is an automated method that uses machine learning to solve real-world problems. It covers the entire process from raw data processing to deployable machine learning models. AutoML is widely considered to be an AI solution to meet the growing demand for machine learning applications. Highly automated AutoML allows non-professionals to use machine learning models and technologies without having to be an expert in the field. By applying the end-to-end automatic machine learning technology, not only can users obtain AI-based solutions and build solutions faster, but most of these solutions can also perform better than manual tuning.

The AutoML framework in Analytics Zoo can automate the process of feature generation, model selection and hyper-parameter tuning. At the same time, the accuracy of the model trained and generated also usually exceeds that of traditional methods or manual tuning measures. As shown in the first group of comparison in the following figure, when the cycle length in the time series data is not particularly regular, there is considerable deviation when traditional methods were applied to predict the time series data, while the prediction value based on Analytics Zoo AutoML is highly consistent with the actual value. As shown in the second group of comparison in the figure below, there is considerable deviation when traditional methods were used when the peak value in the time series data is not particularly regular, while the Analytics Zoo AutoML model gives higher consistency.

Through engineering cooperation between the Neusoft and Intel teams, Analytics Zoo and its AutoML features have been integrated into Neusoft's RealSight APM platform, and have been applied in numerous time series data scenarios, such as trend prediction, anomaly detection, etc. The platform is at the forefront of the industry and will continue breaking new ground.

Entertainment System - Convenience at Your Fingertips

The new cloud platform for intelligent and connected vehicles has introduced the "Little Whale" in-vehicle OS, jointly created by Neusoft and JD.com. It offers 3 major functions - vehicle to home connection, voice shopping, and intelligent vehicle maintenance. The multi-faceted concept of pedestrian-vehicle-home interconnection is made a reality, bringing warmth to living with IoV.


Vehicle Monitoring

Based on ubiquitous access to public cloud services, a low-latency and highly reliable data channel is established to collect the position and operating status data of vehicles and CAN data in real time. Based on the data lake, real-time monitoring, diagnosis and early warning of vehicle faults become a reality. The centralized management portal can be customized to query vehicle usage, route data history, vehicle logs, etc. On top of that, rapid retrospective analysis of data and customized condition filtering are supported.

The system provides a customized search box, which can flexibly define designated vehicle monitoring, designated model monitoring, vehicle monitoring in designated regions, etc. The underlying machine big data platform layer is able to support 5-year backtracking of data and PB-level response retrieval of massive data in seconds.

Vehicle Positioning

It supports the display and pinpointing of vehicle positions based on GIS maps, and position information is automatically refreshed as the vehicle moves. The system can generate heat maps, histograms, trend graphs, and statistical graphs to present statistics for vehicle data at the specified location.

Parameter Monitoring

An independent window is provided for real-time monitoring of vehicle CAN data, and the monitoring dashboard supports advanced data retrieval, e.g. association, data linkage analysis, etc.

Battery Monitoring

The total battery voltage, total current, SOC, and the voltage and temperature of each battery cell can be monitored, with related information displayed through a histogram for a quick view of current battery consistency.

Route History

It supports the query of historical route data of vehicles, as well as online playback of running vehicle routes.

Monitoring via Split Screen

It supports layout customization of the monitoring dashboard based on split screen display, such as 2-screen, 4-screen, 9-screen, etc., allowing the simultaneous monitoring of multiple vehicles and vehicle types. Fixed-point tracking of license plates and models is supported as well, and individual screen zooming can be done.

Advanced Query

It supports full visual configuration to define advanced data query, analysis, statistics, slicing, dicing, filtering, multi-dimensional analysis, etc. VIN, license plate, model, batch, city, and other parameters can be defined, and massive monitoring data can be filtered. Query of PB data can be done in seconds; storage capacity can be linearly expanded as needed to meet storage and analysis demands that arise when there is rapidly increasing vehicle monitoring data.

Fault Alarm

In case of vehicle breakdown, real-time light or voice alarm can be activated. Fault alarm query is also supported.

A rich variety of monitoring and visual statistics chart library is available to address various monitoring scenarios, providing hundreds of special monitoring charts in multiple categories, such as timing, topology, and warnings, meeting all forms of monitoring needs arising from the O&M scenarios of various applications. Dashboard hierarchical structure can be defined, root problems can be pinpointed quickly, and multi-level dashboard association is supported. At the same time, centralized monitoring and drill-down analysis of downtime data can be carried out.

Flexible configuration of data analysis strategy is provided, and charts can be modified with zero coding, DIY exploration of massive O&M data can be done via formula configuration and drill-down, drill-through, slicing, and linkage analysis can be done with one click. The real-time monitoring dashboard and statistical analysis report allow O&M personnel to generate a shared link instantly when an issue is discovered, this facilitates the sharing of data on the issue in discussion groups, email and WeChat to support problem investigation.

Obtaining vehicle operation data can help the analysis and ascertainment of the causes of vehicle failure and traffic incidents. This is helpful in providing failure warnings, offering vehicle user experience, avoiding major incidents and reducing after-sales maintenance costs. Product and service differentiation can be achieved by transforming from just an automobile manufacturer into one that is also a service provider, using the "manufacture + services" business model.

Vehicle to Home Interconnection

Through the 4G/5G network, the vehicle and smart home can be bridged, and the smart driving journey begins, breaking down the barrier between the vehicle and home. Drivers are able to remotely query and control what goes on at home while in the vehicle. Voice, touch, and many other methods can be used while in a vehicle to access the "Little Whale" platform at home.

Drivers can customize the "going home" mode and "leaving home" mode and switch between modes instantaneously. Users can turn on or off home devices with just a touch. Life is made better with more convenience and comfort.

By setting geo-fencing, associated scenarios can be automatically triggered when returning or leaving home, removing the hassles of manual triggering and bringing convenience and efficiency.

Voice Shopping

Through the voice portal, drivers can do voice shopping while in the vehicle. With JD Express's fast delivery guarantee, purchases are rushed to the home or vehicle.

Intelligent Repair and Maintenance

In accordance with the current status of a vehicle and maintenance plan, maintenance reminders are sent to vehicle owners and economical maintenance packages are recommended. At the same time JD will present all vehicle maintenance businesses under JD's automobile club. From maintenance reminder to online purchase of maintenance products, order placement, and servicing, an integrated, all-in-one service is provided.

Committed to creating reliable intelligent technology for users, the cloud platform for intelligent and connected vehicles, jointly built by JD, Neusoft and Intel, is a shining example of collaboration among 3 industry giants, each contributing its technical expertise and ecological resources.

In urban commuting, amid dense traffic, people are most concerned about reaching their destinations both efficiently and conveniently. JD, Neusoft and Intel has worked together closely to create a cloud platform for intelligent and connected vehicles, bringing about a new dimension of interaction - vehicle-to-pedestrian, vehicle-to-vehicle, vehicle-to-control center - making commute smart, and making everything possible.

Contact Us


Back Top