Since introducing its Autopilot driver-assistance system, Tesla has sought customer consent to collect vehicle data to enhance the system’s performance. With a recent update to the Autopilot software, Tesla is broadening its data collection to encompass video footage.
As the update rolled out, which aligns vehicles manufactured since October 2016 with older models in terms of features, Tesla contacted owners to gain approval for a new data-sharing policy. This message, shared with Electrek, requested permission to “capture short video clips from the car’s external cameras,” emphasizing that this initiative would expedite the realization of self-driving capabilities.
According to Tesla, the video clips would aid in teaching the system to identify elements such as lane markings, street signs, and traffic light positions. The company assured that the collected videos would not be associated with any individual vehicle and maintained that “there is no way to search our system for clips that are linked to a specific car.” Tesla also mentioned that some data may be shared with partners, though it would be done in a manner that preserves anonymity.
Every Tesla electric vehicle manufactured since October 2016 is equipped with a new sensor suite referred to as “Hardware 2.” This upgrade includes more components than the earlier “Hardware 1” setup, featuring eight cameras, twelve ultrasonic sensors, and a forward-facing radar system. CEO Elon Musk has indicated that this infrastructure could support fully-autonomous driving, but upon the launch of Hardware 2, Tesla did not have the software necessary to enable certain Autopilot functionalities that were available on Hardware 1 vehicles, leading to the need for the recent update.
While the collection of video data is a new development, Tesla has been gathering data from its Autopilot users for quite some time. This initiative is part of an effort to enhance Autopilot functionality through a process known as “fleet learning.” If a vehicle successfully navigates a particular route, the pertinent data is shared across the fleet, helping all cars to better manage similar driving conditions. This extensive pool of information is likely to provide Tesla with a competitive edge in the quest to develop fully-autonomous production vehicles.
For self-driving cars to effectively interpret their environment, they require real-world data. Other companies have traditionally relied on fleets of prototype autonomous cars for data acquisition, but some are now adopting a strategy akin to Tesla’s. Automakers like BMW, Nissan, and Volkswagen are collaborating with Mobileye to collect data from driver-assist sensors in consumer vehicles, and a new startup called DeepMap aims to profit from managing data collection for autonomous driving systems.
Mobileye, which previously provided cameras and image-recognition technology for Tesla’s Autopilot, severed ties with the automaker last year. The Hardware 2 sensor suite was developed following the end of the partnership between Tesla and Mobileye.
.
