Shulou(Shulou.com)11/24 Report--

On August 17, Zhixing held a media open day this afternoon and officially released its self-developed high-speed navigation system H-INP (Highway-Idriver+ Navigation Pilot).

It is reported that the system is the first high-level autopilot solution using "heavy perception, light map" technology in China. The pre-production Advanced Auxiliary driving solution also marks the full commercialization of Smart Walker in the passenger car field, which is expected to receive at least two fixed points by the end of the year.

▲ press conference site

In addition, during the event, Zhiwalker also gave a comprehensive and detailed explanation of its brand-commercialization strategy and core product self-driving brain. Based on the self-driving brain, intelligent walkers can effectively solve the problem of low reuse rate of self-driving technology scenarios.

Founded in 2015, Zhiwalker is one of the first entrepreneurs to enter the self-driving industry in China. In the past seven years since its establishment, Zhiwalker has formed the achievement of building a self-driving universal platform with annual revenue of hundreds of millions of yuan. As high-level assisted driving has become a label for smart cars, this market is also ushering in more players, and Smart Walker is one of them.

First, it does not rely on high-precision maps, and the self-developed domain controller is different from other high-speed navigation systems on the market. Zhiwalker's H-INP adopts the technical solution of "heavy perception and light map", which has become the first high-level self-driving solution in China that does not rely on high-precision maps.

According to the wise walker, "heavy perception and light map" means that it does not absolutely rely on high-precision maps, and of course it can be applied when there is a high-precision map. when there is no high-precision map, we can reduce the dependence on high-precision maps through key technologies such as space-time joint multi-task fusion perception scheme and online high-precision vector reconstruction.

▲ Smart Walker Test car

For example, H-INP in expressways, urban expressways, ramps and other scenes can be reconstructed by vector maps, check HD map data errors / not updated in time, and further improve driving safety. In the quasi-high-speed scene (national provincial highway), it can be reconstructed with the help of path navigation and vector map, and can still support automatic driving auxiliary functions such as car-following and automatic lane change. at the same time, drivers can be reminded to pay attention to the status of traffic lights and driving safety at traffic light intersections.

Through the above design, the Zhiwalker H-INP high-speed navigation system can effectively solve the problems of low update frequency and low coverage of high-precision maps. What is more worth mentioning is that H-INP uses a platform design, and a series of products can be developed under this unified architecture to meet the needs of different scenarios. According to Wisdom, the future will also be based on the design scheme to build a city navigation system C-INP system suitable for urban scenes.

Specifically, H-INP can be used in expressways, urban expressways, connecting ramps, parking and other scenes. It has rich functions such as autonomous cruising, off-ramp, autonomous lane change, tunnel passage, intelligent yaw, over-control mode, automatic parking, collision warning, etc., and can realize auxiliary driving in different periods (day / night) and high-density traffic flow. At the same time, H-INP relies on the car shadow model and cloud data closed loop to achieve a rapid iteration of the function.

In the design of the overall system architecture, H-INP has the characteristics of functional modularization, interface abstraction and so on. In the H-INP scheme design, the architecture is layered and decoupled, which makes it more convenient to make use of the latest technological achievements; modular reuse of algorithms, application design, development, integration and debugging through graphical tools, focusing research and development on the algorithm itself; deeply optimized communication and scheduling mechanism, which provides multi-level communication mechanism to adapt to the local characteristics of module deployment, and provides deterministic and CPU-GPU joint scheduling.

In the aspect of hardware architecture configuration, H-INP adopts the vehicle specification-level high computing power domain controller independently developed by Wisdom, and the computing power can reach 128 TOPScore CPU with a computing power of 30K DMIPS. At the same time, H-INP adopts a multi-fusion sensing scheme of 6 cameras and 5 millimeter wave radars to achieve effective cost control.

Second, Q4 will launch a mass production version this year, and it is expected to get orders from two car factories. At the press conference, Intelligent Walker CTO Wang Xiao told Cashi that the H-INP is still in prototype testing and is expected to produce a mass production version in the fourth quarter of this year. And will achieve not less than 200000 km mileage, while accumulating no less than 2 million frames of data.

▲ Wisdom Walker CTO Wang Xiao

According to Wang Xiao, H-INP is expected to fulfill orders from at least two automakers by the end of the year, and Zhiwalker aims to achieve no less than 500000 kilometers of test mileage and no less than 5 million frames of data by the second quarter of next year.

Up to now, the unmanned vehicle of Zhiwalker has safely driven more than 5 million kilometers. At the same time, Zhiwang launched comprehensive cooperation with Huawei and took the lead in creating a national production self-driving solution in the industry, realizing the self-control of the whole scheme and ensuring the security of chips and information.

In addition, Wisdom believes that data and the ability to use data iterative algorithms will be an important basis for maintaining technological advantages. At present, based on the landing of a wide range of products in multiple scenarios, smart walkers have formed a perfect data closed loop covering data collection, data processing, data management, data training, evaluation, etc., vehicle-based reasoning-cloud training.

▲ Smart Walker Test car

Under the background of commercial acceleration, how to quickly adapt and apply self-driving technology to different vehicles in different scenes is the core competitiveness. Therefore, intelligent walkers have innovatively developed a model-based development method, which can effectively solve the problem of low scene reuse rate of self-driving technology.

Up to now, the driverless brain of intelligent walkers has been successfully applied in many types of vehicles such as intelligent travel, intelligent life and special applications.

Conclusion: in recent years, high-speed L2, L2 + and other auxiliary driving products begin to accelerate the boarding of passenger cars, and more and more consumers begin to use this intelligent product. Nowadays, high-level auxiliary driving, including high-speed and urban scenes, is becoming an important competitive point for car companies to "increase the size" of their products.

When smart walkers join the field of high-speed and high-level assisted driving, I believe they also see a promising opportunity behind this market. As more and more players join, it is believed that more competitive products will enter the market.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.