One of the biggest challenges in the popularization of self -propelled vehicles is safety and reliability. To provide safe driving to the user, it is crucial that the autonomous vehicle thoroughly, effectively and effectively monitors and recognizes the environment, as well as safety threats to residents.
One sec Tesla is trying not to issue disconnection data best Other companies were provided to develop autonomous driving systems, a group of Beta Tesla FSD testers reporting data independently for some time.
Based on this limited data set, beta tesla FSD It can travel only a few kilometers between the exclusionWhile other autonomous driving programs, such as Waymo and Cruise, report an average of tens of thousands of miles between closures.
On WaymoOne of the methods used to assess the safety of the driver is testing based on virtual, test and real driving scenarios.
To identify appropriate test scenarios, they use existing driving data from WAMO many years of experience, failure data, such as police accidents databases and failures intercepted by CAMS Dash and specialist knowledge in the design sphere, including geographical areas, driving conditions and types of roads. Over time, Waymo still adds new and representative scenarios that they encounter on public roads and simulations, or when they expand to new territories.
The WAYMO script -based script database, developed since 2016, is based on millions of miles driven on public roads, as well as on thousands of real accidents, and provides comprehensive coverage of dangerous situations. Because the most common types of accidents are similar, no matter where you drive, their database can be used as a base line for each city, enabling faster scalability. It covers a wide range of common situations that can happen almost anywhere, such as a pedestrian crossing against a signal or when the car pulls out of the driveway.
In the last study Published in IEEE Transations of Intelligent Transport Systems, a group of international researchers led by Professor Gwangil Jeong from the Incheon National University in Korea, developed an intelligent comprehensive system comprehensive in real time based on deep learning and specialized in self -service situations.
“We have developed a detection model based on Yolov3, a well -known identification algorithm. The model was first used to detect 2D objects, and then modified for 3D objects”, develops prof. Jeon.
The team fed the collected RGB images and data from the point cloud as input data to Yolov3, which in turn the output classification labels and boxes limiting with confidence results. Then they tested its performance using the Lyft data set. Initial results have shown that Yolov3 has achieved extremely high detection accuracy (> 96%) for both 2D and 3D objects, exceeding other current detection models.
This method can be used for self -propelled cars, autonomous parking, autonomous delivery and future autonomous robots, as well as in applications that require detecting objects and obstacles, tracking and visual location.
