Compiled / Zhu Lin
Editor/ Zhang Linyu
Design / Zhao Haoran
Source: Futurecar, Reuters, by Eric Walz, Hyunjoo Jin
Toyota-owned Woven Planet Holdings, which is developing technology including autonomous driving, is developing technology for Toyota itself, taking a new approach to advancing autonomous driving and not using expensive vehicle sensors like lidar, Reuters reported on April 6.
Woven Planet uses low-cost cameras to complement autopilot systems, the same approach as Tesla.
Tesla CEO Elon Musk has been critical of the use of lidar on self-driving cars. He calls the use of lidar a "fool's errand," even though it's widely used by automakers and dozens of self-driving startups.
Musk argues that lidar is "unnecessary" for autonomous driving, and that Tesla's camera-based fully automated driving (FSD) and Autopilot systems do not use lidar data to navigate. Now, Toyota's Woven Planet will take the same approach to perfecting its software.
In an interview with Reuters, Woven Planet said it was able to use low-cost cameras to collect data and effectively train autonomous driving systems in the same way as using expensive sensors. The team at Woven Planet sees this as a "breakthrough" that could help reduce costs and better expand Toyota's self-driving technology.
Self-driving deep learning algorithms rely on large amounts of training data, which helps them get better over time. This data is collected through on-board sensor kits, which typically include cameras, millimeter-wave radar, and lidar. These systems use real-world data collected from vehicles to establish a computer simulation environment and further refine it in this simulation environment.
Volvo's self-driving system uses lidar technology
The data collected from each sensor is processed by "multi-sensor fusion", that is, the information processing system is multi-level, multi-space information complementary and optimal combination processing, and finally produces a consistent interpretation of the observation environment.
The problem is that collecting all this data is a time-consuming endeavor. Tesla, for example, which collects camera data from a connected fleet of more than 1 million vehicles on the road to improve its self-driving technology, is a very large amount of data.
Woven Planet said that for Toyota, using a large fleet to collect diverse driving data is crucial to developing a powerful self-driving car system. But the cost of doing so is high, and there aren't enough fleet sizes to test and refine for self-driving cars that use expensive sensors.
"We need a lot of data. This data can be collected from a small batch of very expensive self-driving cars, and it is not enough to have a small amount of data. Michael Benisch, vice president of engineering at Woven Planet, said.
"Instead, we're trying to show that we can take advantage of Toyota's ability as a big automaker to get a huge database, but with much lower fidelity." He added.
Benish is the former director of engineering at Level 5, the autonomous driving arm of mobility company Lyft, which was acquired by Woven Planet in April 2021 for $550 million.
Before Toyota acquired Lyft's Level 5 assets, Lyft was collecting data from low-cost cameras installed on partner driver vehicles. The camera data also helps the company improve the high-definition maps used to navigate autonomous vehicles. Lyft found that the simple dashboard camera used by drivers was an ideal way to collect additional data on the vehicle's operation in urban areas.
Camera data includes footage of intersections, cyclists, pedestrians, and other drivers acting when Lyft drivers are out to pick up customers.
Waymo's software is able to accurately detect and observe dozens of other road users while navigating
Lyft uses a combination of 3D computer vision and machine learning to automatically identify traffic objects such as other vehicles, pedestrians, and road signs from the camera's footage.
The cameras used by Woven Planet are reportedly 90% cheaper than previous sensors, and these cameras can be easily installed in fleets.
Woven Planet says much of the data from low-cost cameras improves the performance of its system to a level similar to when training entirely on data from expensive sensors such as lidar.
The camera data also helps give Woven Planet a better understanding of human driving patterns. Using visual positioning technology, the system can better track the true trajectory of human drivers when turning or driving in the lane, and improve accuracy.
Collecting data on human driving patterns from cameras will allow Toyota to create self-driving software that more realistically mimics the way human drivers navigate urban environments.
Lyft was able to use this method to create one of the world's largest datasets of real-world driving scenarios. With the acquisition of Lyft's Level 5 division, the torch has now been passed to Toyota, which will build on top of it.
Lyft's autonomous driving division was acquired by Woven Planet in 2021▼
For commercial autonomous vehicles, such as the Robotaxi, which carries passengers, Toyota plans to use an integrated sensor suite on the vehicle. These vehicles will use common set of sensors, namely cameras, lidar and radar. Woven Planet determined that this was the best and safest way to develop Robotaxi at scale.
Benish said this camera-based self-driving technology, like Musk's highly rated Tesla FSD and Autopilot systems, has every chance of catching up with and surpassing self-driving technology that uses lidar and radar sensors.
Many of the advanced driver assistance systems currently offered by new cars also rely on forward cameras, such as lane keeping assist (LKA), automatic emergency braking (AEB), automatic valet parking, and adaptive cruise control (ACC).
In May 2021, Tesla announced. On Model 3 and Model Y vehicles sold in North America, replace the radar sensor with a camera to better support Tesla's Autopilot system and the latest FSD features.
According to the most recent annual rankings of the Intellectual Property Owners Association (IPO), Toyota won more U.S. patents in 2021 than any other automaker, an industry trade group for owners of patents, trademarks, copyrights and other trade secrets.
A significant portion of these patents were awarded to Woven Planet Group, which was founded in 2018 as an expanded business of toyota research institutes (TRI). The group's mission is to develop autonomous driving and other advanced transportation technologies for Toyota.
Toyota's Woven Plant Group includes four companies, namely Woven Planet Holdings, Woven Core, Woven Alpha and Woven Capital. The four companies are developing autonomous driving technology, robotics, smart city technology and advanced transportation solutions.
This article was originally produced by Automotive Business Review
For reprint or content cooperation, please contact the instructions
Illegal reprints must be investigated
Scan the code to join the reader's WeChat group
Communicate about cars
CO to gasoline, electric car terminator?