Research shows that AI driving algorithm performs poorly in bad weather conditions

According to foreign media reports, the Center of U.S. auto manufacturing recently released a report to artificial intelligence (AI) hot water, the results show that the autonomous vehicle algorithm in bad weather conditions Not performing well. The study, conducted by Michigan State University, found that even light rain or drizzle can interfere with the algorithm of self-driving in-vehicle cameras, which may mean that future autonomous fleets can only be found in Arizona (Arizona). Driving in sunny areas such as California and Florida.

Research at Michigan State University identified the core problem not with the primary sensor-camera used to detect obstacles, but rather the algorithm used to sort computer vision data.

Hayder Radha, a professor of electrical and computer engineering at Michigan State University, said: "When we run such algorithms, our detection function is very degraded. Even very weak rainfall can cause some serious problems, and as the rainfall intensity increases. Even the performance of the most advanced mechanisms can be almost paralyzed."

Researchers at the university found that computer vision algorithms could not detect 20% of objects in light rain, and when the rainfall intensity tripled, the detection failure rate doubled. Seasonal variations such as reduced autumn leaves and cold weather have also affected the sensors of self-driving vehicles such as radars and lidars. Automated driving systems must update high resolution maps to cope with seasonal scene changes.

However, the problems caused by the weather did not dampen investors‘ enthusiasm for investing in the development of autonomous driving system startups. For example, in November last year, Korean automaker Hyundai announced its investment in Israeli startup Allegro. Ai, the startup will apply deep learning to computer vision to achieve car navigation. Allegro. Ai was established in 2016, and completed the A round of financing in April 2018, which was worth 11 million US dollars.

Modern car company venture capital service Hyundai Cradle and the home in Allegro, Tel Aviv, Israel. Ai has established a strategic partnership, but Hyundai Cradle did not disclose the amount of investment, but said that the cooperation will accelerate the deployment of the startup‘s artificial intelligence technology. One of them is autonomous driving technology, which is being developed by Hyundai Motor Company and a growing number of car startups, and has achieved varying degrees of success. Hyundai said that it will be Allegro. Ai‘s computer vision technology is deployed in automotive applications such as navigation and real-time driving decisions.

Other companies are also adopting short-term strategies to deploy autonomous vehicles in a controlled environment. For example, Boston-based Optimus Ride is using NVIDIA‘s Drive AGX Xavier computing platform as part of its driverless fleet service, which covers geofencing areas such as communities, campuses and other designated autopilot areas.