intelligent connected car tester: specializes in "finding faults" in self-driving cars to make them smarter
2024-09-04
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
bashan is conducting road tests on intelligent connected vehicles. photo by beijing news shell finance reporter yu jinmin.
just like training a large model, how to make the "brain" of an autonomous vehicle smarter? at this time, intelligent connected car testers are indispensable, and bashan is one of them.
"a small animal that suddenly jumps out or a floating plastic bag may affect the algorithm decision-making of an autonomous vehicle. recently, we have been allowing the 'brain' of autonomous vehicles to deeply learn the traffic police's command gestures, striving to enable autonomous vehicles to 'understand' the traffic police's commands." at the pony.ai r&d plant center in jiading district, shanghai, bashan and his colleagues introduced some of his recent work experiences to the beijing news shell financial reporter.
bashan told the shell finance reporter that being a smart connected car tester is not only about optimizing and upgrading algorithms, but also a challenge to accurately capture and understand complex traffic environments, flexibly respond to emergencies, and continuously optimize decision-making logic.
recently, the ministry of human resources and social security, the state administration for market regulation, and the national bureau of statistics jointly released 19 new occupations, among which smart connected vehicle testers were prominently listed.
intelligent connected car tester：specializedoorgiveautonomous drivingcar“find differences”
"i spend half of my work in the car and half in front of the computer." this is how bashan describes his daily life. perhaps because he often has to follow the car for testing, his face has a healthy wheat color after being "kissed" by the sun. this is only a very small part of bashan's daily work. more often, he is immersed in sorting and analyzing the massive data and complex scenarios sent back from road tests. after locating the problem, he sends it to the engineers for optimization and solution.
bashan graduated from university in 2016. "i actually studied mining engineering in college. after graduation, i didn't want to work in a coal mine, so i found a job as a programmer and became a 'code farmer'. three years ago, i also saw the prospect of autonomous driving, so i chose to start from '0' and became an autonomous driving car tester." bashan told shell finance reporter.
as an autonomous vehicle tester, his job, in simple terms, is to make autonomous vehicles "smarter" and "learn" to drive safely and efficiently in complex and changing traffic environments.
when talking about the intelligent algorithms of self-driving vehicles, bashan seemed very confident. "in fact, a small animal that suddenly jumped out or a floating plastic bag may affect the decision of the self-driving vehicle." bashan told shell finance reporter that in a large number of daily tests, he and his colleagues found that the "reaction" of self-driving vehicles to plastic bags was relatively "extreme." "in the early days, the vehicle's judgment of plastic bags was not very accurate, and it was not sure whether it could run over them. sometimes it would brake suddenly, which was definitely not good for the user experience." bashan said, "we collected a lot of data, and engineers gradually optimized the obstacle avoidance capability. now self-driving vehicles are very mature in identifying obstacles such as plastic bags and can make more reasonable responses."
"my role is to find faults with self-driving cars," bashan said in summarizing his job.
during the testing of autonomous vehicles, there is a special situation called "long-tail scenarios", which refers to those scenarios that occur very rarely but may pose a challenge to autonomous vehicles once they occur. flying plastic bags are long-tail scenarios, which require the autonomous driving system to have a high degree of adaptability, robustness (system robustness) and emergency handling capabilities. long-tail scenarios are rare and unpredictable, so they are given more attention during the testing process. this requires testers to continuously design and test these "long-tail scenarios" and make autonomous vehicles "smarter".
compared to sitting in front of a computer and looking at the data and parameters sent back by the test vehicle, bashan enjoys the feeling of following the vehicle more, which is more real and intuitive, and he is also willing to encounter "new problems". during a following vehicle test, bashan encountered a puppy crossing the road and was almost run over. because the puppy was short, the vehicle camera was not that sensitive. "in fact, i also like to raise cats and dogs. i feel sorry for the safety of these small animals, but when an autonomous vehicle encounters such a situation, it cannot respond well and needs time to consider."
after returning to the company, bashan conducted repeated tests on scenarios involving small animals, allowing the "brain" of the autonomous vehicle to face small animals more calmly. bashan said that when driving on the road, he often encounters vehicles parked temporarily. "is he going or not? is he stopping or staying? as human drivers, we are not very good at judging." after this problem was fed back to the autonomous vehicle tester, the model was optimized, and more processing experience was "fed" to the autonomous vehicle. now the autonomous vehicle can make judgments in a very short time when encountering such situations.
continuous optimizationpolish every detail，meetingbyordinary passengersidentityexperiencedownowntestofproduct
bashan is debugging the onboard computer of an intelligent connected car. photo by beijing news shell finance reporter yu jinmin.
in every car-following test, the tester is not only an observer and recorder, but also a "coach" of the autonomous vehicle. what really makes autonomous driving technology close to people's hearts and adapt to complex and changing traffic environments are the algorithms and models that are continuously iterated and optimized based on human experience.
during daily testing, bashan pays close attention to the vehicle's physical comfort, such as why sudden braking or acceleration is necessary and whether it can be optimized. "only by polishing every detail can passengers have a better experience."
bashan told the shell finance reporter that he is actually happiest that more and more people are using autonomous driving technology in their lives, whether driving or taking a taxi. "sometimes i will call a car as an ordinary passenger to experience the future travel mode that i have helped to shape, and observe the performance of autonomous driving vehicles with a professional and relaxed attitude."
the "future of jobs report 2023" released by the world economic forum shows that new technologies, digitalization, green transformation, etc. are reshaping the career landscape and also placing higher demands on the skills and qualities of workers.
some industry experts pointed out that technological innovation is not the enemy of employment. in the wave of autonomous driving and intelligent connected vehicles, some new occupations are gradually moving from concepts to reality.
yu jinmin, financial reporter of beijing news shell
editor: yue caizhou
proofread by liu baoqing
