Laziness is the driving force behind the advancement of human science and technology, and the sweeping robot is the product of this logic. But the original sweeping robots were either rotten to death or stupid. Dyson, who had also launched sweeping robots, launched the V6 handheld wireless vacuum cleaner in 2014, and it immediately became an internet celebrity vacuum cleaner. On behalf of consumers at the time, they cast a vote of distrust for sweeping robots.

But what about today, which is approaching 2021? This time we tested the Roborock T7 Pro. As the flagship sweeping robot, it is equipped with a variety of the most advanced civilian technologies and represents the highest level of sweeping robots. Then buy a vacuum cleaner and clean it manually?

Or is it enough to buy a robot cleaner? Can robots replace humans and clean the home?

Powerful cleaning ability

Admittedly, to define whether a cleaning device is good or not, many people think that the key is to clean it. In fact, T7 Pro has probably the most powerful suction power of 2,500 Pa among household sweeping robots. In contrast, the Mijia mopping and sweeping robot, also produced by Stone, has a suction power of 2,100 Pa, and Cobos T8 AIVI is 1,500 Pa.

T7 Pro, even if it is not the best suction machine among the sweeping robots, can be regarded as the top equipment among them, plus the main brush module is available aroundFloating space, close to the ground when cleaning, the effect is better. In the actual test, the strong suction power of T7 Pro is definitely one of the factors that make our impressions unforgettable. Moreover, although its suction power is strong, the relatively low noise will not make people feel too annoying.

Maybe you may think that even if the cleaning ability of the robot cleaner is not good enough, you can still make up for the lack of performance through frequent counting. However, many people may have children and pets in their homes, and it is impossible to smoke them several times a day. And on weekends, you don’t want the sweeping robot to be at work multiple times, so that it will interrupt your rest. So for the sweeping robot, strong suction is still very important.

But to be more critical, its biggest weakness in cleaning is the use of a single-sided brush design (above). Although the single side brush can reduce the chance of being stalked by debris, it will also reduce the coverage when vacuuming. Therefore, T7 Pro must use a more complex and dense cleaning path when cleaning, in order to completely cover the entire house and suck the corners clean. Although this does not make the T7 Pro’s cleaning coverage smaller, it does require more cleaning time than its double-sided sweeping robot.

T7 Pro introduces a new constant voltage controlled water tank. Picture source: Rock Technology.

In addition, in the past, many users often complained about the small water tank capacity of the Roborock sweeping robot, which could not meet the single complete mopping needs of large houses. Therefore, T7 Pro Introduced a new constant voltage controlled water tank (above), and repaired the short board of mopping ability through the diaphragm pump scheme.

But to be honest, this new water tank is only repairing the original old shortcomings. The mopping ability can only meet general needs. It does not have a powerful high-frequency brush like competing flagship sweeping robots. Ground components, or equipped with water tanks that can automatically change water, have not greatly improved the mopping ability. In contrast, although the T7 Pro’s mopping function is not bad, it still cannot match its flagship identity.

The sweeping robot has long eyes

But if you say whether a cleaning device is good or not, it’s not true if you only look at the cleaning ability. Are we going to use a vacuum cleaner? Isn’t it because we don’t want to use mops? To be honest, “laziness” is the driving force behind the advancement of human science and technology, so the performance of the robot cleaner should not be focused on:

How to work with the least amount of physical effort for the greatest cleaning effect?

So, our focus is still on how the T7 Pro can free users’ hands and how to clean our home without anyone operating. As the flagship T7 Pro, it basically already has most of the high-end sweeping robot functions on the market, including LDS laser sensor, cliff sensor, intelligent partition function, restricted area setting and so on.

But the real biggest selling point of Rock T7 Pro is the combination of LDS laser sensor and binocular camera.

Although recent sweeping robots have also used LDS laser sensors for obstacle avoidance and real-time positioning, since most of the laser sensors used by sweeping robots are 2D lidars, they can only generate flat maps. Still can’t see those obstacles that are above or below the field of view (above).

Therefore, although these robots can avoid obstacles, they must not have too many debris on the ground when performing cleaning tasks, otherwise these sweeping robots will eat your slippers, wrap the wires, and get stuck on the threshold. Many users have asked the sweeping robot to automatically clean when they are not at home. What they see when they return home is not a spotless home, but the sweeping robot entangled with wires and immobile, shoes pushed into the depths of the cabinet, And the pet feces dragged across the house.

This Stone T7 Pro, in addition to the LDS laser sensor, also uses binocular vision technology commonly used in aerial drones. The parallax generated by two 120-degree cameras can generate a wide range of depth information. Small obstacles such as shoes and power strips that were originally difficult to detect with LDS laser sensors have become another obstacle avoidance redundant system besides laser radar. According to Roborock, the success rate of obstacle avoidance in the same complex environment is increased by 90%, thereby reducing the embarrassing situation of the sweeping robot being stuck again.

However, the biggest weakness of binocular vision technology is that it must be operated in an environment with enough light. Partial sweeping robots often need to get into the table or the bottom of the cabinet. It is inevitable to work in a dark environment, and obstacle avoidance is likely to occur. The problem of reduced capacity. Therefore, T7 Pro has also added an infrared fill light (above), which can maintain a certain degree of obstacle avoidance even in an environment with insufficient brightness under the bed, which is a fairly complete solution.

Image recognition and avoidance function

In addition, since T7 Pro has added visual technology, it also has new functions for image recognition through artificial intelligence. According to Stone Technology, T7 Pro can quickly identify common obstacles in the home such as wire clusters, patch panels, shoes, bar stool bases, and weight scales in real time (260ms/time). Through image recognition, T7 Pro can adopt different safe distance avoidance strategies for different objects: For example, T7 Pro will judge the cable group as “very dangerous” and avoid obstacles early at about 5cm, while slippers are “safe”, only about 3cm. Start to avoid obstacles.

Because the T7 Pro only has a forward-facing binocular camera lens, it can only achieve one-way vision avoidance. But because it can record all obstacles encountered in the map at the same time (above), this allows T7 Pro to achieve a full range of avoidance capabilities, and in principle, it can avoid various obstacles.

But under our actual measurement, the Rock T7 Pro cannot avoid all obstacles, and not all objects can be identified. Even if the object to be avoided can be identified, it does not mean that the avoidance is successful. What is the reason for the failure to avoid? It’s hard to say at the moment.

But the author estimates that this is related to the principle of vision itself. Since binocular vision and image recognition all need to extract the feature points of the picture, but the author’s floor is pearly white, dark objects will have more obvious characteristics, while light objects are relatively difficult to identify (above) ), this may affect the recognition ability of T7 Pro to some extent