Lyft Makes Self-Driving Research Public in Bid to Speed Development
Lyft is publicly releasing the Level 5 Dataset, a subset of its self-driving research, in the hope that the data will expedite further innovation and encourage collaboration in the autonomous driving sphere. The Level 5 Dataset—which Lyft calls “the largest publicly released dataset of its kind”—is segmented into training, validation, and testing sets, and includes over 55,000 human-labeled 3D annotated frames, data from seven cameras and up to three lidar sensors, a drivable surface map, and a high-definition spatial semantic map. Lyft Level 5, the company’s autonomous driving division, plans to release more data as their research continues.
In releasing the Level 5 Dataset Lyft is likely looking to advance several goals. Lyft may be hoping the widespread dissemination of their dataset will gain their methods and configurations greater traction within the AV sphere. Moreover, by publishing, Lyft opens the data’s accompanying challenges and troubleshooting to academics and other researchers, who might further develop the data and effectively do Lyft’s work for them, gratis. The release doubles as a recruiting tool, as researchers who successfully develop Lyft’s data will likely receive attention from Lyft’s onboarding team; to this end, Lyft is holding a research competition for individuals to work on the dataset, with cash and the potential to interview for Lyft’s team among the prizes.
51VR initiated China’s first autonomous driving simulation bluebook
Chinese virtual reality and artificial intelligence company 51VR has published China’s first self-driving simulation bluebook, a reference document containing statistics and information. 51VR, alongside several Chinese companies and institutions, released “Annual Research Report on Autonomous Vehicle Simulation in China (2019)” at the 6th International Congress of Intelligent and Connected Vehicles Technology (CICV) in Beijing. The report harnesses the opinions of Chinese industry experts and details the development status of China’s AV simulation test. The publication introduces 51Sim-One, China’s first complete automated driving simulator and testing platform, and supposedly addresses all aspects of AV simulation testing, from the test’s significance to software development and testing standards.
51VR hopes to use the bluebook as a launchpad to create an AV supply chain complex including OEMs, suppliers, startups, and testing agencies. The collaboration that went into the report—eight institutions jointly issued the bluebook, not including 51VR—indicates Chinese AV companies may be imitating their American counterparts, which have been sharing data and working together to hasten AVs’ arrival.
Daimler and Bosch’s driverless parking gets OK to operate without human supervision
German regulators have given Daimler and Bosch the green light to operate their automated valet parking feature without a human safety driver, marking the first time a fully driverless level-4 automated parking program has been approved for public use. Visitors of the Mercedes-Benz Museum in Stuttgart, Germany, will be able to order their Mercedes-Benz autonomous car to park itself with a smartphone app. The museum’s garage is equipped with Bosch sensors that will guide the vehicle to a parking space; the car will return to the drop-off point when the driver comes back.
Germany amended a road traffic bill to include autonomous vehicles in 2017, allowing drivers to transfer control to the car on public roads, but stopping short of allowing fully autonomous, driverless vehicles. Germany still lacks state regulation and approval processes for fully self-driving cars, so Daimler and Bosch collaborated with local transportation officials since beginning work on the project in 2015. Their cooperation is indicative of an environment where regulation and the technology are emerging side by side, each influencing and informing the development of the other.
Man injured by self-driving shuttle in Salt Lake City; Driverless Bus Hits Pedestrian in Vienna
A series of accidents involving self-driving shuttles in the past week have drawn existing issues with autonomous vehicles into sharp relief. In Salt Lake City, a self-driving shuttle being tested by the Utah Department of Transportation (UDOT) and the Utah Transit Authority was driving in the 1950 West business park when it suddenly stopped, causing a 76-year-old man to fall from his seat and sustain facial injuries. In response, UDOT lowered the vehicle’s speed and said a host on board the vehicle would warn riders about abrupt stops, although it has reaffirmed it believes the shuttle is safe.
In Vienna, Austria, an autonomous bus by AV startup Navya bumped a 30-year-old woman’s knee while driving at 7.5 miles per hour, prompting city officials to temporarily halt tests until the conclusion of their investigation. Navya began testing a pair of autonomous buses in June, and it is unclear when tests will resume. These incidents should caution both AV developers and the general public and serve as a reminder that although the technology is moving forward, it is not yet at a stage where the vehicles are sufficiently safe and immune to error.