Founders Securities published a research report stating that the curve of progress for advanced intelligent driving technology is becoming increasingly steep. Although the experience of end products and the commercialization process have not yet broken through the threshold (passenger car ADAS is still far from freeing drivers' hands, and the penetration rate of commercial applications such as Robotaxi and port unmanned driving is limited), the progress of technological advancement is clearly accelerating, and advanced intelligent driving is expected to accelerate towards L4 evolution.
The to-B nature of Robotaxi determines that there will be phased differences between it and single-vehicle intelligence in terms of technical routes, BOM costs, and generalization capabilities. The team is optimistic about the future reliance on algorithmic capabilities to bridge the gap in hardware BOM investment. Vehicle-road collaboration helps to raise the upper limit of the system capabilities of single-vehicle intelligence, but it does not affect the penetration of advanced intelligent driving in passenger cars.
Advertisement
The curve of progress for advanced intelligent driving technology is becoming increasingly steep: In 2021, Tesla expanded the perception field through BEV → In 2022, the occupancy network solved the recognition of general obstacles → In 2023, the end-to-end framework solved the iteration efficiency and scene generalization issues from rule-driven to data-driven → In 2024, the industry began to integrate the LLM framework into the intelligent driving algorithm framework (simultaneously solving corner-case cognition, visualizing AI decision-making processes, enhancing trust in human-computer interaction, and enhancing the interpretability of decision-making results).
Although the experience of end products and the commercialization process have not yet broken through the threshold (passenger car ADAS is still far from freeing drivers' hands, and the penetration rate of commercial applications such as Robotaxi and port unmanned driving is limited), the progress of technological advancement is clearly accelerating, and advanced intelligent driving is expected to accelerate towards L4 evolution.
Clarify the relationship between Robotaxi, single-vehicle intelligence, and vehicle-road collaboration: 1) The to-B nature of Robotaxi determines that there will be phased differences between it and single-vehicle intelligence in terms of technical routes, BOM costs, and generalization capabilities: Robotaxi does not have the possibility of passengers actively taking over the vehicle, and it needs to be responsible for supervision, requiring more regulatory code and hardware redundancy (cross-city generalization is more difficult and the entire vehicle BOM cost is higher); passenger car L3/L4 pursues full-scene intelligent driving at the cost of mass-produced vehicles, currently lagging behind Robotaxi in terms of take-over rate indicators, but the algorithm iteration speed is faster (thanks to end-to-end and data accumulation), and it is optimistic about the future reliance on algorithmic capabilities to bridge the gap in hardware BOM investment; 2) Vehicle-road collaboration helps to raise the upper limit of the system capabilities of single-vehicle intelligence, but it does not affect the penetration of advanced intelligent driving in passenger cars: The long-term goal of passenger car OEMs in planning advanced intelligent driving models is to achieve L3/L4 nationwide, and the level of road-side intelligence in some cities/some sections does not affect the planning of car companies.
Risk warnings: 1) The risk of intensified industry competition; 2) The risk that the improvement of the experience of advanced intelligent driving/penetration rate increase is not as expected; 3) The risk that the opening process on the regulatory side is not as expected; 4) Geopolitical risks.