Internet of Vehicles Forum | When will vehicle-road collaborative autonomous driving arrive?

03. When will vehicle-road collaborative autonomous driving arrive?

The mass deployment of vehicle-road cooperative autonomous driving requires three prerequisites:

(1) As single-vehicle intelligent technology develops to an advanced stage, autonomous driving companies have a clear understanding of the necessity of vehicle-road collaboration. The development of autonomous driving technology also conforms to the 80-20 principle. The initial 20% investment can solve 80% of the problems, but the remaining 20% ​​of the problems require 80% investment to solve.

Therefore, in the early stages of autonomous driving development, the improvement of a single vehicle's own capabilities is sufficient to cope with the use of many working conditions, and autonomous driving solution providers are not enthusiastic about vehicle-road collaboration. Only when the development of autonomous driving has advanced to a certain stage and the aforementioned cost issues, blind spot perception issues, traffic light acquisition issues, and route conflict issues can no longer be solved through single-vehicle intelligence, will autonomous driving companies have enough motivation to introduce vehicle-road collaboration technology. This is also the reason why leading autonomous driving companies such as Baidu and Yushi are the first to adopt vehicle-road collaboration technology to supplement single-vehicle intelligence.

(2) The source of vehicle-road collaboration information reaches a sufficiently accurate level. The essence of vehicle-road collaboration is to allow vehicles to obtain more information to assist in decision-making. The impact of erroneous information may be fatal.

Traditional traffic informatization does not require high accuracy and recall rate of perception, and has no requirements for position accuracy. In the era of vehicle-road collaboration, especially when serving autonomous driving, the accuracy requirements for roadside perception must be raised to the same level as single-vehicle perception. As more and more autonomous driving solution providers enter the field of vehicle-road collaboration, related technologies are making rapid breakthroughs.

Traffic light status acquisition is another rigid requirement for autonomous driving. Vehicle-road cooperative information transmission itself will hardly cause errors in status analysis, but the signal machines themselves vary in quality, resulting in the information source not being 100% reliable. Due to technical and management limitations, signal machines in many areas do not provide status output interfaces, so many manufacturers have developed signal machine learners in order to adaptively acquire the status and periodic changes of traffic lights. However, signal machine learners have a learning time, and at some intersections that use adaptive periodic control, it is difficult for the learner to track the rapid changes of signal machines. To fundamentally solve this problem, it is necessary to thoroughly transform the signal machines to allow accurate information to be obtained from the source.

(3) The density of autonomous vehicles on the road has reached a certain level. Autonomous vehicles exist as production tools. People expect to fully realize the commercial value of autonomous driving technology by replacing manual labor, operating around the clock, and reducing the probability of errors.

In the early days of technology application, the density of autonomous driving vehicles was not high, and there were enough road resources for autonomous driving vehicles to use, so the demand for vehicle-to-vehicle collaboration was not high. With the continuous deployment of autonomous driving vehicles, the number of vehicles that need to be carried on a unit of road resources has increased. In order to make more effective use of limited road resources, it is necessary to introduce vehicle-road collaboration. The practice of autonomous driving in ports and logistics parks has already illustrated this. In the early days of the project, the intelligence of a single vehicle was sufficient to cope with the operation of one or two lines. With the continuous opening of new lines, conflicts between vehicles are increasing. Autonomous driving companies are seeking to use vehicle-road collaboration to improve the efficiency of park operations.