Since the advent of the era of electrification, automatic assisted driving has been highly concerned by the industry. At present, the dispute between the pure visual route and the perception fusion route in perception technology has become a common topic, and the recent decision of Toyota has brought this issue into the public eye.
Recently, Toyota has publicly stated that the company will use low-cost cameras to collect data and develop and upgrade its autonomous driving system through effective training. This approach helps reduce costs and scale up technology. This means that Toyota has officially joined the "visual faction" camp to which Tesla belongs.
Once Toyota was also a supporter of the "radar faction"
Lidar was originally proposed by Toyota and Nvidia, and large companies such as Google, Volkswagen, Daimler, Bosch, and Baidu were all supporters of the "visual" camp. We all know that the "visual faction" and the "radar faction" are not high or low in essence, but they have their own advantages and disadvantages.
The essence of the pure visual autonomous driving route is to obtain information about the surrounding environment through high frame rate, high-resolution cameras, but the complex terrain and road conditions will cause huge computing pressure on the visual perception system, and the visual perception ability will be greatly reduced in bad weather. Companies that insist on lidar believe that the mixed use of multiple sensors is obviously more computationally powerful, more flexible and more accurate. However, the disadvantages of lidar are also obvious, that is, the application cost is high, the requirements for computing power are high, and the threshold is too high, which leads to the inability of some car companies to apply at this stage.
Back in 2016, Toyota announced a $1 billion fortune to set up a division in Silicon Valley called the Toyota Research Institute (TRI). The most notable thing about Toyota's self-driving prototype is its 7 lidars from velodyne. But today, Toyota believes that if only a small amount of data is collected from a small number of more expensive self-driving cars, it is not enough to support the development of self-driving technology.
Finely deducted costs, which is very Toyota
Obviously, the first factor in Toyota's consideration is in terms of cost. At present, the price of lidar in the world is still high, and the price of lidar has been lowered in the mainland because of the existence of companies such as Huawei and DJI. But in foreign countries, in 2021, the price of mainstream lidar on the market is basically still around $2,000, while the cost of a single camera of Tesla Model 3 was only $65. This is obviously difficult for Toyota, which strictly controls costs, to accept.
At present, although Tesla is very single in hardware, but relying on the collected data and computing power, Tesla FSD has achieved good results so far, so will Toyota's technology be stronger than Tesla?' It's hard to say that Toyota's annual global sales of more than 7 million vehicles can be the best source of data. And for price-based reasons, lidar is not suitable for large-scale widespread promotion, unless the cost can be reduced to the same price as the camera. However, if that day really comes, presumably a higher-spec perception system has already replaced the status of lidar.
Write at the end
On the whole, there are indeed very obvious advantages and disadvantages of autonomous driving technology such as Tesla FSD with only a single hardware, but for the current Toyota, such a development idea can indeed quickly solve the problem of its own advanced automatic driving technology.