Author: Bai Yang
Highway navigation assisted driving should be the fastest growing intelligent driving function in the next 3 to 5 years, without a doubt. It is expected that from 2023, at least 15 new cars will be equipped with the highway navigation assisted driving function annually.
However, at the same time, the highway navigation assisted driving function is also currently the “most dangerous” intelligent driving function, and almost all high speed Pilot functions that have been pushed out have had accidents of varying degrees.
One reason is that the high-speed scenario itself is relatively dangerous, with very low redundancy for errors, and the other reason is that most consumers are not clear about the boundary of the highway navigation assisted driving function. Most car companies also do not have much motivation to do user education in this area.
Therefore, before the explosive growth of the navigation assisted driving function, we believe that more valuable content must be created to help everyone have a deeper understanding of the navigation assisted driving function. This is also the reason why our AutoLab has done the “Highway Navigation Assisted Driving Comprehensive Evaluation” for two consecutive years.
Although it is called the “Highway Navigation Assisted Driving Comprehensive Evaluation”, in fact, the “Comprehensive Evaluation” is to “trick” everyone into coming, because I know that everyone likes to watch exciting films. Essentially, we want to:
Not compare who is higher or lower, but test the ability boundary.
How to Test the Highway Navigation Assisted Driving Function?
How to quickly understand the ability boundary of a product with limited time and energy investment? We have jointly developed a set of test standards for highway road testing and scene restoration with 36Kr, GeeksPark, ClassStar Channel, Super Charging Station, XEV Research Institute, and Dr. Zhang Kangkang.
First of all, for the high-speed road test part, in order to control the environmental variables as much as possible, we chose 8:00 on November 4th, with 8 test car models (consistent system settings and route settings) all starting from Shanghai, passing through 512km of highway, and using the navigation assisted driving function throughout the journey, arriving at the destination.
In addition, we will also set up a “human driving control group” this time, which will start at the same time on the same route to record the process time, facilitating comparison with the system time consumption of the 8 tested car models.
Throughout the process, we will ensure safety first and let the system handle acceleration, deceleration, and lane changing control as much as possible. We will record data from multiple dimensions, including time consumption, lane changing tests, lane changing success rate, merge-in/merge-out success rate, system degradation times, and human takeover times/reasons.
To unify the threshold for drivers to take over the assistive driving, we have conducted driver training one week in advance. We have tried out the actual road functions and defined the dangerous scenarios in a standard manner, so as to align the judgments on dangerous road scenarios among the drivers as much as possible and reduce the impact of personal differences on the data.
In addition to the daytime test, we will also have a nighttime test on the same route to facilitate understanding of the stability changes resulting from differences in light. Therefore, the total length of our high-speed road test this time is about 1024km.
Of course, high-speed road testing with controlled variables alone is not enough. Many extreme scenarios are the most dangerous, so we have boldly taken over the Bosch Donghai testing ground, planning to use safe testing equipment to replicate some very dangerous scenarios.
For example, vehicles involved in accidents slanted on the road, construction scenarios, mixed scenarios of people, cars, posts, and barrels, high beam headlights encountering stationary vehicles at night, and irregular obstacles, etc. We will showcase the actual performance of different car models in some high-speed dangerous scenarios that have occurred before by replicating them.
If you have any testing suggestions for the scenario replication part, you can scan the code to enter the group and give feedback to optimize this testing together, to “eliminate” the dangerous scenarios that may exist in the navigation assistive driving and increase everyone’s awareness of risk prevention.We also hope that through this effort, we can help more car enthusiasts understand the “ability boundaries” of different vehicle navigation and assisted driving functions at the current stage, and enjoy the easy and safe experience brought by the advancement of intelligent driving technology.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.