Many companies from around the world exhibited at " Automotive World, " an exhibition of cutting-edge automotive technologies held at Tokyo Big Sight from January 22 (Wed.) to 24 (Fri.), 2025, with themes such as electric vehicles, automated driving, and carbon neutrality. Automotive World 2025" is composed of several exhibitions, including the Car Electronics Technology Expo and the Sustainable Technology Expo, but the 8th Automated Driving EXPO, an exhibition specializing in automated driving, was held in Shenzhen, China, DeepRoute.ai, a technology company specializing in the research, development, and application of smart driving solutions, headquartered in Shenzhen, China, will be exhibiting for the first time! We reported on the seminar held by Mr. Xuan Liu!
Technology session by Mr. Xuan Liu!
A technology session by "Xuan Liu" of DeepRoute.ai was held at the venue. DeepRoute.ai is a unicorn company that has developed the first mass-producible smart driving solution in China and has received investment from Mercedes-Benz.
DeepRoute.ai will begin road testing the E2E model in 2024, and is also releasing a model with this model to the market. Unlike conventional end-systems, the E2E architecture aggregates multiple modules by connecting them in a single network. The aggregation avoids information loss between modules, which leads to higher quality to the learned rules, according to the company.
Initially, this functionality was only available in four cities in China, but within three months, the scope of use was expanded to include all of China. The system has been demonstrated in 31 provinces in China and has accumulated 23,000 km of driving data. The data collected can be used to optimize models and systems, and is expected to further improve the ability to respond to various complex situations in the future.
At present, there are more than 30,000 models equipped with DeepRoute IO, and more high-quality data is continuously coming in. The data acquired is being used to create a solid foundation for the development of the next generation of VLA models (Vision, Language, Action).
The E2E model has the disadvantage of being uninterpretable, and the current model has not always been able to respond well to unusual road signs or scenes that are not often seen. However, the addition of the language allows for precise behavior and can handle more complex scenes. It also has the ability to identify road signs and analyze which lane to use, even in reversible lanes, as well as the ability to make decisions to understand its surroundings, such as surrounding vehicles and scooters, and adjust its driving.
The VLA model, which is expected to be in mass production by mid-2025, is expected to make automated driving decisions easier to understand, and will be a quantum leap in immediate efficiency and reliability.
The first model equipped with the E2E model of DeepRoute.ai self-driving technology will be released in August 2024, and 30,000 cars have already been sold. The model has tripled in sales compared to the non-Smart Driving model, making it the best-selling model in its class. In the four months from September to December 2024, it captured 10% of the NOA market share in urban areas in China. By the end of 2025, sales of models equipped with DeepRoute.ai's self-driving technology are expected to exceed 200,000 units, and the company plans to launch projects for more than 10 models, including SUVs, MPVs, and off-road vehicles.
DeepRoute.ai, which recently announced a strategic partnership with German automaker smart, plans to establish an office in Germany in 2025 to aggressively expand its business overseas, and is planning to start road tests. It has been revealed that the company plans to establish an office in Germany in 2025 and begin road tests. The company's goal is to provide smart driving vehicles not only to mainland China but also to the rest of the world in cooperation with global automakers. In addition, the company is aiming to incorporate its self-driving solution, which is currently focused on automobiles, into other robots in the future.
Expect a future where time spent traveling by car can also be used effectively!
At the DeepRoute.ai booth, an experimental video of a model equipped with DeepRoute IO was shown. The quality of the self-driving technology, such as recognition of obstacles and smooth movement at complex intersections, was evident in the self-driving video shown to the public. In the past, travel time on trains and buses could be effectively utilized, but in the future, when traveling by car, the driver may be able to leave the driving to the automated driver and make effective use of his/her travel time by playing games, reading, sleeping, etc. This time, the technology was demonstrated for the first time in the Japanese market. This is the first time that this technology has been demonstrated in the Japanese market, and it will be interesting to see if cars equipped with "DeepRoute.ai" smart driving technology will appear in Japan in the near future!
Sorry, this article is only available in Japanese and 繁體中文. at the moment.
