Guangzhou Automobile Science and Technology Day： Exposing Solid-State Battery／Intelligent Driving Technology without Map

[Interpretation of car home’s Fresh Technology] GAC TECH DAY 2024 was held grandly today. Compared with previous technology days, this event not only invited local media, but also invited a large number of foreign media to show the global audience the strong technical strength of China’s first-tier car companies. On this year’s Science and Technology Day, GAC released technologies such as all-solid-state batteries and pure visual intelligent driving without drawing. The media at the scene also have the opportunity to experience the city NDA navigation assistance function that will be listed soon and carried on Haobo HT. Let’s take a look at the details of these forward-looking technologies!

● Map-free pure visual intelligent driving technology

The theme of this Guangzhou Automobile Science and Technology Day is "Science and Technology Vision-See the World with New Vision". This theme is obviously to introduce GAC’s brand-new GARCIA intelligent driving system. According to the official, this new system is the first unmanned pure visual automatic driving system in China and will be fully mass-produced in the spring of 2026.

Driving without a plan has become a new industry trend, and all major first-line automobile brands in China are involved in this field. Intelligent driving without map, as its name implies, does not rely on high-precision map data, but only uses various environmental sensors equipped with vehicles to build maps, plan routes and make vehicle control decisions in real time. Abandoning high-precision maps is conducive to promoting high-order intelligent driving systems in different places more quickly.

The pure vision scheme is to get rid of sensors such as lidar, and only use the camera to identify obstacles around the vehicle, which provides reference for vehicle control decision-making for the intelligent driving system. The latest intelligent driving system of Tesla, which we are familiar with, adopts the pure vision scheme.

To realize the intelligent driving system without graph and pure vision, a powerful algorithm is needed as a technical endorsement. Guangzhou Automobile Research Institute has accumulated a certain amount of technology in this field, and relevant algorithms and technologies have been verified in some international competitions, winning the first place in the pure visual list of nuScenes Autopilot Test Competition in August 2022 and the first place in Argoverse 2 Motion Prediction Challenge in July 2023.

GARCIA’s intelligent driving system can not only perceive regular vehicles, pedestrians, riders and ice cream cones, but also identify irregular obstacles such as carts and animals.

Generally speaking, the environmental identification algorithm of GARCIA Intelligent Driving System has not opened a very big gap with the products of the first echelon of Intelligent Driving, and its toughness seems to be more reflected in information decision-making and execution.

Compared with the current mainstream design of intelligent driving system, GARCIA’s multi-target tracking and forecasting ability is expected to achieve stronger and better route planning ability. GARCIA intelligent driving system has learned a lot of driving data from experienced drivers, and is also familiar with the traffic regulations, which is expected to make the intelligent driving system achieve higher traffic efficiency while complying with the traffic regulations.

At present, the intelligent driving technology mostly calculates the optimal speed under a given route or the optimal route under a given speed. Facing the complex and changeable road environment, there is still much room for improvement in traffic game ability. The spatio-temporal joint planning technology can effectively improve the traffic game ability of intelligent driving system, thus achieving higher traffic efficiency and more comfortable ride experience.

Of course, the spatio-temporal joint planning technology needs to be used together with the ability of target tracking and prediction to have the effect of "1+1>2". The powerful traffic game ability brought by GARCIA’s intelligent driving system is worth looking forward to.

The former automatically labels the image data learned by the intelligent driving system, which greatly improves the efficiency and reduces the cost compared with manual labeling. The latter uses a generative AI model such as Stable Diffusion to generate diverse image data for the intelligent driving system to learn, which improves the recognition ability and accuracy of the system.

According to the planning of GAC, GARCIA’s intelligent driving system is still about two years away from mass production. What will be the field of intelligent driving after two years? I don’t think anyone can accurately judge, but it is foreseeable that some advanced designs of GARCIA’s intelligent driving system are made for the next round of intelligent driving competition. I think competitors must have laid out relevant technologies in private, so only time can give an answer.