Jia Haonan sent from the co-driving temple smart car reference | Official account AI4Auto
After rolling the price, rolling up the battery life, rolling up the large screen and rolling the radar...... In 2024, how else can smart cars roll?
Most likely, it is a city pilot, which is more advanced and complex than high-speed pilot, which can independently identify traffic lights and complex road conditions in the city, and send you to your destination according to the navigation route.
In fact, the trend is already clear. At the end of 2023, the city pilot "Kaicheng" battle is in full swing, and in terms of configuration, the hardware solution (with lidar) that supports the corresponding functions is even more standard for models in the early 200,000s.
In the first year of the popularization of city piloting, the heroes rose together, asking the world, Xiaopeng, and ideal......
In the mighty wave of intelligent driving "involution", an unexpected "dark horse" also ran out.
At a time when many car companies in the industry are breaking through single intelligent driving functions and scenarios one by one, GAC, which is deeply hidden in the field of intelligence, is about to push the self-developed urban NDA (Navigated Driving Assist) function to users.
What's more, GAC has taken the lead in completing the on-the-road test of the L3 autonomous driving "hands-off-eyes" function, and has realized the integration and interoperability of GAC's ADiGO PILOT in all scenarios, redefining a new standard for intelligent driving function experience.
To understand how this "dark horse" is made, the most direct way is to find it from the source of technology research and development - Xu Wei, director of the intelligent driving technology department of the GAC Research Institute, chatted with qubits.
The intelligent driving racing pattern and the number one dark horse in the first year
In the urban pilot race, the basic pattern is judged by whether the city pilot is really pushed to users to realize intelligent driving in a variety of complex scenarios. At present, there are only a few companies in the group that have done it, and some in the industry either have insufficient scene coverage or single functions.
This is enough to show the difficulty and threshold of mass production of urban pilots.
On the one hand, the city navigation function has just entered the mass production stage, and no one has yet been able to achieve perfection in terms of experience and coverage scenarios.
In addition, the actual scale of running is not large, and the iteration speed has not yet taken off, which also makes the racing pattern and power ranking change at any time.
It is in this context that GAC has carried out its "debut" of intelligent driving ability, and the road test has chosen a round-trip route between the GAC Research Institute located in the suburbs of Guangzhou and the GAC Center located in the core urban area, with a total distance of 58 kilometers, and the test time is the rush hour of 9 am in Guangzhou.
Smart car reference
, Likes236
In addition to the regular ability to go straight at traffic lights, merge onto ramps, and make autonomous left/right turns, the performance of GAC City's NDA has several impressive highlights.
First of all, for the temporary construction road conditions, the system can immediately identify the passable range of the road and avoid it reasonably.
The second highlight is that the system successfully handles the challenge of being left and being given another car: first brake to avoid it, and then pass the intersection without hesitation after confirming that it is safe:
There is also a complex game scene that is also surprising, on urban and rural roads, first pass through an unprotected right turn with irregular lane markings:
Immediately after the turn, the oncoming large vehicles occupy the road, and at the same time, it is also accompanied by the challenges of forcibly congested trucks and cars, low-speed two-wheeled vehicles passing in the opposite direction, and actively getting on the high-speed ramp.
From the perspective of intelligent driving strength, GAC not only keeps up with the rhythm of the first echelon, but will soon push the urban NDA function, and does a better job in the coverage of corner cases: urban NDA has the ability to deal with most difficult scenarios, easily does not quit the takeover, the system is more robust, and the user experience is more smooth and coherent.
Of course, the bigger highlight is GAC's L3 autonomous driving function, TJP (Traffic Jam Pilot), which runs in actual road conditions, realizing the function of "taking off your hands and eyes".
It is completely legal and compliant, because GAC is also one of the first companies in China to obtain a L3 autonomous driving road test license. Moreover, GAC's self-developed L3 autonomous driving function is completely "real and usable".
The key is in the trigger and exit mechanism: when the road conditions are met, the system can activate the Level 3 autonomous driving function - the TJP traffic jam navigation system. Under the TJP function, the vehicle can automatically carry out lane centering, speed keeping, vehicle distance keeping, etc., and the driver can operate the car machine and play with the mobile phone "without his hands and eyes".
Once the vehicle in front changes lanes and leaves, the system loses its following target, or the car is about to make a turn or enter or exit the ramp, the system can switch to the L2++ level of pilot assistance driving function by itself, and the vehicle and instrument will remind the user to hold the steering wheel and look ahead.
The meaning of "real usability" lies in the fact that the L3 function is not separate, it can be seamlessly switched with the L2++ function, so that GAC's intelligent driving function can be integrated and interconnected on the road in all scenarios.
Compared with other new power players, GAC, which has many years of car manufacturing precipitation, has a larger volume and system, although it is not easy to turn around, but once the turn is completed, the momentum will be unstoppable.
At the beginning of 2024, when the smart car track was still arguing about the truth and falsehood, and who would get the L3 license first, GAC proved itself to be the most noteworthy "number one dark horse" in the involution of intelligent driving.
What is the strength of GAC's self-developed intelligent driving technology?
The R&D task of GAC Intelligent Driving is undertaken by GAC Research Institute, the hub of GAC Group's R&D system.
Here are two more words about the path and method taken by OEMs in the intelligent transformation.
Similar to GAC, the main R&D task is undertaken by the Central Research Institute, and then adapted to the models of various sub-brands, which is called "centralized" intelligent R&D.
The other path is the opposite of "distributed" intelligent R&D, which means that the research institutes and sub-brands within the group build their own teams and "race horses" with each other.
It is difficult to conclude which R&D model is more efficient and has a better product experience, but more and more car companies, after several attempts and explorations, have chosen the "centralized" model.
The advantages of this are evident here at the GAC Research Institute.
Taking GAC's urban NDA experience as an example, there are two biggest features, one is the robustness of the system mentioned earlier, and the other is the integration of all scenarios of intelligent driving, and the integration and interconnection between functions. They correspond to GAC's specific technology accumulation at the "micro" level and the construction of the technology system at the "macro" level.
GAC's intelligent driving system is based on BEV visual perception, combined with the reasoning ability of Transformer, and introduces many innovations in perception, regulation and control.
For example, in order to ensure higher safety and reliability in full-scene intelligent driving, GAC adopts a multi-sensor redundancy solution with three lidars to achieve no blind spot perception coverage within 200 meters around the vehicle, and at the same time greatly improves the perception effect in harsh environments such as haze, rain and snow, and dark night.
For example, at the regulatory level, the traditional spatio-temporal separation scheme is difficult to handle scenarios that require horizontal and vertical coordination, such as accelerated overtaking, lane change within small GAP, and lateral dodging under emergency cut-in.
For true "all-scene" intelligent driving, GAC has a completely different technical system. According to Xu Wei of GAC Research Institute, the core reason behind the "dark horse phenomenon" of GAC intelligent driving is the adoption of a set of platform-based technology system, "no detours, one step in place".
Now the common rhythm of intelligent driving popularization is to first go on a single cruise control, lane keeping and other functions, then on the high-speed pilot, and finally on the city pilot. Different functions have inconsistent requirements for algorithm capabilities, underlying computing power, and sensors, which has caused OEMs to repeatedly reconstruct the technical system back and forth, which is also the fundamental reason why most car companies cannot connect in terms of intelligent experience.
GAC R&D Institute has built a set of technical systems from the beginning, whether it is urban NDA, high-speed NDA, or L3 autonomous driving, or even the L4 Robotaxi demonstration operation fleet that is completely independently developed and built by L4 unmanned driving, all of them share a set of platform-based software solutions, including infrastructure computing power and basic algorithm modules.
Therefore, GAC's high-speed NDA, urban NDA, and even L3 autonomous driving can be integrated and interconnected.
Dark horses are not black
In the first half of the new wave of the automobile industry, there is the "GAC phenomenon" that cannot be ignored: in the "electrification" transformation competition, GAC has become one of the most mature and successful car companies among the old car companies.
Even in terms of absolute sales figures, Aion has dominated the list for a long time, and GAC is also the only old OEM that can be on an equal footing with the new forces.
Xu Wei told Qubit that in the second half of intelligence, GAC even started earlier than electrification.
Since 2010, GAC R&D Center has started to develop its own ADAS solution based on single millimeter wave radar.
In the next two years or so, GAC began to carry out forward-looking layout research and development of "unmanned driving", and released GAC's first-generation unmanned concept car WitStar. This early layout result was directly transformed into the L4 Robotaxi autonomous driving demonstration operation business jointly built by GAC Research Institute and Ruqi Mobility.
In 2017, shortly after the release of the "Intelligent Connected Vehicle Technology Roadmap", GAC Research Institute established the Intelligent Driving Technology Department in a timely manner based on the development plan of intelligent driving technology, focusing on building its own forward R&D system based on basic capabilities and product application scenarios, and establishing a unified technology base and R&D process for functions and experience through independent R&D or cooperative ecology.
For example, through the L4 demonstration operation test for Robotaxi, GAC has accumulated a large number of corner cases in urban areas, and through the background "shadow mode", it can continuously learn and compare with the driving behavior of human drivers, and collect valuable measured data at high density. All of these can greatly supplement the problems of single data and weak scenario generalization in the L2/L2++ field, and also make GAC's self-developed technology stack have the genes of urban NDA very early.
The premise of all this is still that GAC has opened up the integrated technology system of L2, L3 and L4 many years ago, unlike many players who spend a lot of money to build different technical systems at different stages.
GAC has become the "number one dark horse" of intelligent driving in 2024, running into the first echelon of independent brand intelligent driving, and leading the transformation of old car companies...... Of course, there are advantages of latecomers, but the core is still to understand the essence of autonomous driving technology, and work backwards from the use scenario to build a complete set of technical systems.
Therefore, the "GAC mode" of intelligent driving should be: a set of platform-based technology system, a process-based underlying base, and a series of efficient self-developed tool chains.
This also makes the view that has been popular for many years have a factual argument for the first time: the volume and scale of the old car companies are there, and once they are intelligent, they will catch up with or even crush them in minutes.
The progress and strength shown by GAC come from GAC's engineering thinking and ability to build cars for many years, as well as top-down innovation genes, and an intelligent team that truly understands AI and scenarios.
This is the most critical and unique connotation of the "GAC model".