According to TrendForce's analysis, autonomous driving is a key application direction of Edge-AI. Driven by the end-to-end model boom set off by Tesla, all parties are accelerating the deployment of AI technology and computing power. It is expected that by 2025, other automakers will start mass production of end-to-end architectures, but they will mainly choose modular end-to-end models with more advantages in interpretability and debugging. This kind of model is data-driven and highly dependent on diverse data, and generative AI, with its openness and creativity, can be used to generate data in multiple and rare scenarios to help model training and effectively solve the long-tail problem of data. With the gradual improvement of the regulatory environment, Level 4 autonomous driving autonomous taxis are expected to accelerate the process of scenario replication and commercial operation. However, whether it is electrification or autonomous driving technology, in the process of development, geopolitical factors may make technology and business expansion more challenging.
According to TrendForce, 2025 will usher in a new era of development for the autonomous driving industry, with the mass production of modular end-to-end models and the commercialization of Level 4 robotaxis.
Modular end-to-end model mass production
Technological progress and advantages
Fusion of multimodal large models: The latest evolution direction of end-to-end models is to deeply integrate multimodal large models, such as Vision-Language-Action Model (VLA). The VLA model has higher scene inference and generalization capabilities, can better cope with complex traffic scenarios, and has a stronger ability to understand the world end-to-end, which is expected to become a key springboard for the leap from L2 assisted driving to L4 autonomous driving.
Increase the upper limit of performance: Compared with the traditional sub-modular scheme, the one-piece end-to-end model eliminates many artificially set priors, and when the data volume is large, it will not be constrained by a priori, so as to further increase the upper limit of the system. For example, Momenta's end-to-end solution has surpassed its level of intelligent driving by dozens of times in a short period of time by removing the a priori.
Simulate human learning logic: In order to solve the problem of low end-to-end solutions, some enterprises have adopted human-learning-like logic. For example, Momenta's end-to-end model is divided into short-term memory and long-term memory, and new data first enters short-term memory, and then enters long-term memory learning after verification, which can not only quickly adapt to new situations, but also ensure the stability and reliability of the model.
Figure: Modular/non-end-to-end autonomous driving system architecture
Level 4 Commercialization of Robotaxis Accelerates
Market potential and prospects
Global market size growth: The global autonomous driving market size continues to grow, with the global autonomous driving market size of about $158.3 billion in 2023, a year-on-year increase of 29.97%, and is expected to grow to $273.8 billion in 2025. This provides a broad market space for the commercialization of Level 4 Robotape.
The rapid development of the Chinese market: China's autonomous driving market is in a stage of rapid development, with the scale of China's autonomous driving market reaching 330.1 billion yuan in 2023, a year-on-year increase of 14.1%, and is expected to approach 450 billion yuan in 2025. With the continuous maturity of technology and the reduction of costs, the commercialization process of Level 4 Robotaxi in the Chinese market is expected to accelerate.
On the other hand, autonomous driving car companies are also accelerating their layout, such as Baidu Apollo, Pony.ai, WeRide and other companies have made significant progress in autonomous driving algorithms, high-precision maps, sensor technology and other aspects. These companies have increased their investment in the field of Level 4 robotaxis to promote their commercial application.
The close cooperation between industries such as automobile manufacturing, Internet technology, and communication operations has promoted the R&D, testing, and application of Level 4 robotaxis technology. Through the sharing of resources and the exchange of resources, various industries work together to overcome technical problems and accelerate the commercialization process.
It is reported that in 2025, the on-board operation safety risk management and control system will be gradually deployed and applied in the autonomous driving system to serve the supervision and implementation of autonomous driving operation safety. This will further enhance the safety and reliability of Level 4 Robotaxi and promote its large-scale commercial operation.
Related:
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (1)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (2)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (3)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (4)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (5)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (6)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (7)
Top 10 Key Technology Market Trends for 2025: A Forecast Analysis (8)
