As we navigate through the transformative landscape of the automotive industry, it is increasingly clear that we have entered an era dominated not just by electric vehicles but by an unprecedented wave of intelligence embedded within our carsThis shift transcends mere electrification; we are witnessing a rapid evolution where even the most basic automobiles are becoming interconnected and intelligent computational devices, redefining the driving experience as we know it.

During the recent Daqingshan Technology Conference for Intelligent Vehicles 2024, a prominent figure, Zhang Yongwei, who serves as the Vice Chairman and Secretary-General of the China Electric Vehicle Hundred Committee, articulated this sentiment robustlyHe stated that the automotive sector is experiencing advancements in intelligent technology at a pace that could soon outstrip the developments seen in the electric vehicle sector

Within the next couple of years, it is anticipated that cars lacking intelligent features will be a rarity.

Supporting this assertion, data released by China's Ministry of Industry and Information Technology revealed that the penetration rate for new passenger vehicles equipped with Level 2 autonomous driving capabilities or higher reached an impressive 55.7% in the first half of this yearNotably, vehicles with Navigation On Autonomous (NOA) capabilities are beginning to mark their presence in the market, achieving an 11% penetrationProjections indicate that by the end of this year, sales of smart connected vehicles in China could surpass 17 million, with an overall penetration rate exceeding 60%. Such statistics reflect the remarkable investments that automakers are making into intelligent technologies, fostering the widespread application of advanced driving systems and large model technologies.

However, alongside this progress arises a crucial issue—autonomous driving systems are reliant on vast amounts of data

• Ensuring Reasonable Returns on Renewable Energy Generation
• Micron Technology's Plunge
• AI Drives Hua Bei Credit Limits
• Strategies for Federal Reserve Rate Cuts
• In the New Era, Turn to AI for Guidance

These demanding requirements pose significant challenges for automotive companies in ChinaGathering, processing, and analyzing data on a grand scale has proven difficult, and the complexities involved include difficulties in achieving extensive data collection, insufficient coverage of long-tail scenarios, and spiraling costs that accompany data acquisition.

Industry experts have reached a consensus that promoting data sharing could be the vital key to resolving these challengesBy facilitating an environment where data can be collaboratively shared among different entities, organizations can significantly lower their data acquisition costs while enhancing the speed and efficiency of data processing.

As the focal point of competition shifts towards intelligent driving, the last couple of years have seen artificial intelligence and large model technologies burgeon, propelling the development of smart connected vehicles from rules-based systems to data-driven models

In this paradigm shift, high-quality data has emerged as an essential element in training sophisticated algorithms that underpin intelligent driving systems.

Nevertheless, the challenge remains: obtaining high-quality intelligent driving data is far from straightforwardAt the same conference, Li Bin, the Chief Technology Officer of Beijing Kaiwang Data Technology Co., articulated the contrasting positions of companies like Tesla, which commands a fleet of over two million vehicles globally capable of comprehensive data collection across diverse driving scenariosBy contrast, domestic manufacturers face significant limitations with their own data collection efforts, often restricted to specific areas and driving conditionsThe requirement for relevant qualifications further complicates these efforts, making it cumbersome for China’s autonomous driving data initiatives to scale effectively.

As demand for data continues to increase exponentially alongside the growth of autonomous capabilities, companies find themselves bracketed by rising costs

For instance, the collection of 400,000 data frames might necessitate an extended timeframe of approximately 40 weeks for data acquisition and another 80 weeks for processing—a cumulative expense that could exceed several million yuanLi suggested that if companies could successfully implement a data-sharing model, with part of the data collected independently and another portion sourced from partner firms or suppliers, it could drastically reduce both collection times and costs, thus accelerating progress in model training.

Not only would wider access to data streamline operational costs, but it would also enhance the frequency at which autonomous driving models can be trainedSwiftly improving the iteration process consequently allows for a richer, more effective training regimen for these vehicles.

At present, many companies within the automotive industry perceive data as a core asset, leading to a reluctance to share, which exacerbates the phenomenon of “data silos” within the sector

Nevertheless, diverse datasets are critical for training and optimizing models in autonomous driving systemsBy embracing data sharing, companies can minimize redundant collection processes, broaden the diversity of sample data, and overcome the long tail of data scarcity that inhibits model performance in rare or extreme driving situations.

Diverse viewpoints are emerging that advocate for the establishment of an open cooperative ecosystem through data shared in order to push smart driving technologies towards maturity, ultimately facilitating large-scale implementation across global markets.

The dialogue surrounding data sharing and collaboration is already in motionLiu Bo, the Vice President of Zero Data Technology in Shanghai, emphasized the importance of recognizing data as a product of diverse agents operating in varied environments over timeIt is through meaningful sharing that a second growth curve can be nurtured.

“Previously, data primarily circulated within the confines of a single organization, transferring between different departments

While this internal circulation has already shown efficiencies, it is only a starting pointNow is the time to cultivate external circulation that traverses organizational boundaries and crosses industry verticalsSuch an approach will ultimately create a second growth curve,” said Liu.

To facilitate this transformation, building a robust automotive big data interaction platform is vitalTackling challenges such as data alterability, ownership issues, uncontrolled circulation, privacy leaks, and opaque calculations requires targeted actionLiu underscored the need for a foundational infrastructure to address these pain points effectively.

During the conference, Huo Jingyu, General Manager of the East District at Four-Dimensional Map New Technology, provided insights into the rapidly increasing data demands stemming from the heightened perceptual capabilities of smart driving

The staggering statistics indicate that each vehicle can generate over 150 parameters every few seconds and that data volumes could reach up to 5 terabytes per hourDespite the vast potential, the sensitive nature of autonomous driving data poses considerable apprehensions for companies; the lengthy and convoluted chain from data collection to processing adds further layers of complexity.

In response to these challenges, innovative infrastructure for artificial intelligence is being heralded as a viable solutionHuo detailed how his company has developed a closed-loop processing system encompassing data from the vehicle, the road, and the cloudThis system is designed to filter out useful data that can enhance the training of autonomous driving systems.

“This framework enables automakers to access precise, actionable data needed for training systemsWe’re keen to offer our specialized algorithms to automakers for integration, cultivating a collaborative environment that fuels innovation and enhances capabilities,” Huo remarked.

Kaiwang Data portrays itself as a connector facilitating seamless data flow robustly

They have partnered with an affiliate of Beijing Yizhuang Intelligent City Research Institute to launch China's first collaborative platform for "vehicle-road-cloud data," which has now begun operational deployment.

This platform is dedicated to gathering intelligent traffic data from various sources, providing robust, secure applications for smart data usageIt not only delivers finely tuned data but also the GPU computing clusters essential for model training.

The fluid circulation of automotive data assets has the potential to dissolve the existing silos stagnant in the realm of autonomous driving.

As Zhang Yongwei articulated, the pathway toward smart development must be characterized by openness, collaboration, and a concentrated focus among companiesDeveloping intelligent vehicles cannot be a solo endeavor; success hinges on the ability to foster an open ecosystem