For autonomous vehicles, it is important to detect and perceive the surrounding environment, and excellent sensors can help cars make the right choices. At present, the dispute between pure visual route and perception fusion route in perception technology is an old topic, and it is also a choice problem faced by every company when developing autonomous driving technology.
In recent years, both traditional car companies and new car-making forces have bet on lidar as a perception solution, and only Tesla has insisted on the camera solution, and now Toyota motor has also joined the ranks. Recently, Toyota's subsidiary Woven Planet, which develops self-driving technology, said it is trying to advance autonomous driving technology without the use of expensive sensors such as lidar, which can collect data with low-cost cameras and effectively train its self-driving system.
Opt for a more pragmatic camera, which is very Toyota
At present, the sensors used in self-driving cars mainly include cameras, millimeter wave radar, ultrasonic radar, lidar, GPS and so on. Each of these sensors has advantages and disadvantages, and they also complement each other. Why would Toyota risk falling behind the existing self-driving system and opting for cameras, you might ask? Once you've seen their pros and cons, you might be able to understand this approach.
From the perspective of advantages and disadvantages, the advantage of lidar is that it has high accuracy, long detection distance, and can achieve higher accuracy than the camera in low-light environments. Therefore, models equipped with lidar can more accurately obtain the three-dimensional information of the target, and the resolution is high and the anti-jamming ability is strong. But at the same time, the development cost will be higher.
The camera wins in the lower cost, which undoubtedly reduces the cost of automatic driving, so that it is expected to be truly borne by ordinary consumers. At the same time, the camera is smaller and easier to integrate into the car design, without making the vehicle shape too abrupt. However, the camera also has its own weaknesses. For example, in the dark environment, perception is limited, accuracy and safety have decreased; in extreme weather, or other situations that cause the camera to not work (such as the lens is obscured, strong backlight), it is necessary to rely on powerful algorithms to ensure the completion of perception processing, and the upper limit of the automatic driving ability that can be achieved is low.
At present, the perception ability of lidar is slightly stronger than that of cameras, but Tesla and Toyota have chosen to focus on cameras, which may be related to the high cost of lidar. The head of Woven Planet said that equipping vehicles with expensive sensors is not a sustainable approach, and the large amount of data collected by cameras has great flexibility, which can expand richer functions on a certain hardware cost.
Shifting from lidar to pure visual autonomous driving, Toyota is in a hurry
In fact, whether it is a pure visual route represented by a camera or a perceptual fusion route represented by lidar, it represents the attempts of different explorers to achieve a path for autonomous driving. Vehicles equipped with lidar, high-precision maps, etc., although the hardware advantages are more obvious, but the cost is certainly high. The high cost is bound to limit its development space and is not conducive to large-scale rollout.
However, although the cost of using a camera will be lower, relying solely on cameras to achieve full automatic driving requires a large amount of vehicle driving data to create a computational model to make automatic driving more intelligent. From this point of view, Toyota and Tesla still have some way to go if they want to achieve higher-level automatic driving.
At present, the biggest criticism of lidar is the cost problem, as the cost of lidar has dropped again and again, models equipped with lidar have also sprung up. Up to now, great wall, Xiaopeng, Jihu, GAC and others have clearly decided to mass-produce models equipped with lidar, and the first model equipped with lidar, the Xiaopeng P5, has been officially launched.
While other brands are embracing lidar, Toyota's choice of cameras also has its own considerations. As we all know, the development of autonomous driving systems requires a lot of data, and Toyota is currently slightly behind in the autonomous driving track, in order to quickly catch up with the mainstream team, the choice of reliable and low-cost cameras does help to increase the collection speed. In addition, the pure visual method based on the camera is difficult first, it is difficult at the beginning, and the advantages behind it will become more and more obvious, and Toyota is more willing to do difficult things.
It is worth mentioning that when Toyota shifts to a pure vision route, it does not mean that it has completely abandoned the road of multi-sensor fusion. For safety reasons, other kinds of self-driving cars such as Toyota's driverless taxis will still use sensors such as lidar.
Whether it is a lidar or a camera solution, in fact, there are advantages and disadvantages, as for who can become the mainstream in the future, it also depends on who breaks through the obstacles first. But it is undeniable that as a traditional veteran car company, Toyota has obvious advantages in terms of production and sales scale and technical reserves. If Toyota's plan to collect large amounts of data at a low cost goes ahead, it is expected to help Toyota quickly stand in the first array of autonomous driving technology.