On September 3, the "Sub-Forum of 2022 World Artificial Intelligence Congress: AI Business Landing Forum," guided by the Organizing Committee of the World Artificial Intelligence Congress and hosted by EqualOcean, took place in the Shanghai EXPO Center.

As the driving force of productivity in the digital economy era, AI contributes significant incremental value through innovation and ongoing breakthroughs in industrial application scale. Since 2021, the size of AI enterprises has continued to grow, with many listings achieved successfully. Now, companies in the growth and maturity stages outnumber startups, becoming the market's mainstream. This shift signals a new era in AI development. In this market environment, AI enterprises must leverage their core technologies to create economic and social value, which will be the key to their stable survival and sustainable growth.

During this conference, EqualOcean released the "2022 China AI Business Research Report" and the "Top 100 2022 China AI Business Landing Enterprises." Westwell was listed for the fifth consecutive year. Dr. Sun Zuolei, SVP and Qomolo Partner from Westwell, was invited to this forum. With rich R&D and mass production experience in autonomous driving and intelligent robotics, Dr. Sun drew the audience's attention. Currently, he is primarily responsible for the engineering mass production of the company's autonomous driving system architecture and its foundational components.

Part of Essentials of the Speech:

It's a great honor to share with you our progress and practices in the intelligent upgrading of container logistics under the axis of 'carbon peaking and carbon neutrality.' After years of development, Westwell has grown into a sizable team with 70% engaged in research and development. We've been recognized as a national high-tech enterprise. Our vision is to use artificial intelligence technology, paired with autonomous driving technology, to drive the intelligent upgrading of production factors in the global logistics field.

Just last month, China Gov set a short-term goal of 'carbon peaking and carbon neutrality,' aiming for 40% clean-energy vehicles and a reduction of 25% and 20% of carbon emissions from passenger vehicles and commercial vehicles respectively by 2030. The plan emphasizes reducing heavy truck carbon emissions. Based on extensive research and experiments, we believe that electrification is the most feasible method for decarbonizing heavy trucks. According to data released by Cornell, sales of new energy heavy trucks in China will increase by 247% over the next three years, and the market penetration rate of electric heavy trucks will reach 18% by 2030.

Given the high labor costs and significant driver gap, traditional fuel-heavy trucks used in high-frequency start-stop short-distance container transportation are primed for transformation and replacement in the surge of intelligence, automation, and greening. Since establishing our business, we've focused on the intelligent upgrading of the container logistics industry, targeting the ports as an entry point and aiming to improve operation efficiency and operational cost for users through artificial intelligence.

To support this strategy, we have developed WellOcean and Qomolo. This endeavor includes designing autonomous driving vehicles, pure electric vehicles, containers, handling platforms, autonomous driving container shuttles, and our electric autonomous commercial vehicle, Q-Truck.

In 2017, our journey with autonomous driving commercial vehicles began as an exploration to restructure traditional fuel container trucks by wire. We soon realized that if we wanted to exploit autonomous driving's full potential, the vehicle's base form must be custom-made.

Following this understanding, we launched the cab-less electric autonomous commercial vehicle Q-Truck in 2018. Over the past four years, Q-Truck has undergone several modifications to meet specific customer scenario needs and continue improving product performance and user experience based on the integration of autonomous driving software and hardware. This year, as another innovative and bold practice in line with this product concept, we released the 2022 Q-Truck for battery swap, achieving full automation of the process in 5 minutes.

Since entering the B2B field, customer demand has also been shifting as the market evolves. Nevertheless, we maintain the principle that pure technology does not create productivity; only technology integrated into production results in real productivity. Hence, we've been seeking different scenario landing spots and persistently attempting to address users' pain points.

As industrial users require autonomous fleets, not just autonomous driving vehicles, that can work cohesively with their business processes and technological status, we have offered users a comprehensive set of solutions, including autonomous driving. Based on their unique scenarios and process characteristics, we provide tailored and integrated solutions from local to global. We develop and deploy from the ground-level wire restructure and autonomous driving components to the fleet management system and the automated operation and maintenance tool set on the upper layer.

Besides scenario-specific development, we also continue investing in base-level research, staying committed to observing and improving infrastructure R&D. With multiple autonomous product lines, we strive for more efficient development iterations. Hence, Q-Pilot, a self-developed autonomous driving component by Westwell, was conceived. It includes the modules of map construction, positioning, perception, similar planning of Robotaxi in open scenarios, and also supports alignment and interaction with vertical transportation devices in user scenarios. Q-Pilot has been fully launched with all vehicle types in Westwell's full product line, like the 2022 battery swap version of Q-Truck, autonomous shuttle, pure electric container transport platform, autonomous minibus, and the types released last year.

Q-Pilot supports semantic crowd-sourcing with composition and positioning, and the environment built by multiple autonomous driving vehicles to extract semantic information from environmental elements. For instance, in the port scenario, containers in the yard are identified and marked as map atomic elements, synchronously shared with other vehicles in a autonomous driving fleet, to realize the synthesis, distribution, and update of the feature model. This approach enhances the positioning performance by taking advantage of the clustering capability of autonomous driving fleets.

In the latest Q-Pilot release, it integrates a perception engine, converging on multi-task, multi-data sources, and multi-market. This engine supports data preprocessing of 8 high wire harness lasers and 8 high-resolution cameras and provides feature extraction at different scales to facilitate feature projection using deep neural networks. Additionally, it introduces persistent correlation between different sensor sources to address the issues of shielding and observation loss among multiple obstacles.

When the entire engine reasons, different tasks including obstacle segmentation, detection, and prediction share the same resources to preserve computing efficiency and facilitate deployment.

However, learning-based perception methods are sensitive and driven by data, and data engineering is often more challenging than the development of a deep neural network. Given the heterogeneous traffic state of our autonomous driving fleet, we strive to create and adapt perception capabilities on open roads. We continually invest in our data platform, initially setting up a data feedback loop system driven by MLOps, including Continuous Training (CT), Continuous Integration (CI), and Continuous Deployment (CD).

Similar to how we design our vehicles independently, we believe that a combination of software and hardware can provide the best user experience. Hence, we've insisted on pursuing in-house development for hardware. Accordingly, our industrial binoculars have undergone a new iteration that supports PPS time synchronization to facilitate the output of high-accuracy environmental depth information.

Q-Pilot's chassis control has always been equipped with the AVCU developed by our company. Together with our steer-by-wire QWireVCU, it evolved to version 3.0 this year, offering new support for automotive-grade duplicate supplies.

To reduce costs, accelerate deployment, and enable broader scenarios, we initiated research and development on an autonomous driving simulation platform three years ago, namely, WellSIM. This year, we released the official version of WellSIM2.0. It supports the simulation of autonomous driving and the simulation of user operation systems, other operational equipment in user scenarios, and interaction processes between itself and the autonomous driving vehicle. Additionally, it can simulate different weather scenarios and sensors, such as lasers, vision sensors, and diverse traffic flows. It allows the configuration of different opponent vehicles or static obstacles through a traditional scenario device to generate sensor data and test the perception, planning, and decision-making capabilities of autonomous driving to improve the efficiency of firmware iteration in cooperation with Q-Pilot.

Thanks to our continuous investment in technology R&D, our products and systems have served more than 160 group customers across 18 countries and regions since our introduction to ports based on the intelligent cargo handling product line in 2016. In the Tianjin Port, we have aided the commercial operation of customer-end zero-carbon autonomous driving terminals on fleet management systems and single-vehicle autonomous driving technology.

As the benchmark of the first intelligent upgrade of traditional terminals in China, we've provided a horizontal transportation system built upon our autonomous driving pure electric transport platform for the Hairun Terminal of Xiamen Port's intelligent transformation. It has thus become the first terminal in China to ensure simultaneous production and intelligent transformation.

The Q-Truck fleet at the CSP Abu Dhabi Terminal in the Middle East has adopted a brand-new operation model and has generated revenue based on the number of containers. So far, this has led to a 75% decrease in manpower investment by our customers.

We have assisted the internationally renowned port equipment manufacturer, ZPMC, in producing autonomous driving shuttles that have seen commercial operation at the CTN Terminal of the joint port in Stockholm, Sweden.

Q-Truck fleet has carried out handling commercial containers in Laem Chabang Port of Thailand for nearly two years, with complete mix operation of manned inner trucks and outer trucks entering the terminal regardless of operation areas and lanes.

The Q-Truck fleet has been operational in Thailand for 25 months, having operated 531 commercial ships and completing 155,000 TEUs of commercial handling. This has led to a 27.5% annual efficiency improvement for users.

These actions are our concrete steps towards implementing the “carbon peaking and carbon neutrality” strategy. Through intelligent practices and new energy business landings based on autonomous driving products, we are connecting transport capacity to vertical ecosystems to cover a larger operational area.

Deeply rooted in seaport scenarios, we leverage the accumulated autonomous driving vehicles and maneuver capacity and extend them to other container logistics closed scenarios. We implement carbon management of all autonomous driving products launched by Westwell throughout their entire lifecycle. This is supported by the lean management of carbon emissions from our self-developed fleet management operation platform and the energy-saving strategies in vehicle chassis and intelligent driving algorithms.

To further improve user experience, we independently designed PowerOnair—an intelligent battery swap platform for commercial vehicles based on vehicles, battery swap stations, and cloud platforms. This platform features a multi-sensor integration technology that achieves the first side battery swap operation of heavy trucks within 5 minutes through the separation of vehicles and batteries. As such, it entirely automates battery swapping.

To meet the logistics requirements between user stations, we also launched an intelligent internet-connected heavy truck, E-Truck, designed to support trunk logistics in open scenarios. This allows for seamless switching of intelligent driving from closed to open scenarios, supporting our vision of deeply integrating intelligence and green solutions into major container logistics scenarios.

After years of dedicated effort, we're proud to announce that this year, the entire series of Westwell's new energy autonomous driving products are expected to reduce carbon emissions by 2,193 tons. This is equivalent to planting trees over 4,622 mu. We hope Westwell becomes the driving force behind the intelligent upgrade of the traditional logistics industry—with vehicles as carriers, electricity as the core, scenarios as support, and transport capacity as output.