The Roborock T7 Pro was officially released today. It upgraded the AI binocular visual obstacle avoidance and applied the technology of drone and automatic driving to the obstacle avoidance of the sweeper. Its dual 5-megapixel camera can obtain more accurate information on objects, thereby improving its obstacle avoidance effect, and can also avoid pet cakes.
The Roborock T7 Pro Vacuum Cleaner is equipped with the new RR Mason 7.0 stone robot system algorithm. Its “real-time data management algorithm” is responsible for managing and distributing the environmental data collected by the sensors. It is handed over to the “perception and mapping algorithm” for data fusion and analysis, and then the “navigation and motion algorithm” performs dynamic navigation and planning based on the analysis results.
Under the dual power of CPU GPU, when the robot shakes when encountering obstacles, the algorithm will trigger the map anti-shake mechanism to ensure that the map remains stable. At the same time, when the environment in the home changes, the T7 Pro will dynamically update the map to clean it, without bumping or leaking.
In terms of performance, the Roborock T7 Pro uses the same brushless fan as the T7. In Max mode, the suction power of the fan can reach 2500Pa, the built-in 5200mAh high energy density lithium battery, single battery life is about 2.5 hours, and it can clean about 250 square meters of house area.
The constant voltage controlled water tank equipped with the Roborock T7 Pro has a large capacity of 297ml. In the “small water” mode, it can drag the floor up to 250 square meters at a time.
T7 Pro Design
In appearance, the Roborock T7 Pro continues to maintain a modular design and appearance that is basically consistent with the T7 Pro but has made some fine-tuning in detail to make the human-computer interaction part of the entire sweeping robot more reasonable. Let’s take a look together.
The packaging is always a stone style. The hard corrugated cardboard box can protect the machine well, and the simple printing of the line drawing and the name can distinguish the different models of machines easily and clearly.
Take out the body, the appearance of the three physical buttons has changed compared to the past, the top surface has not changed much.
The raised LDS laser sensor will quickly rotate when the cleaning robot starts, scanning the surrounding terrain and obstacle information. At the same time, you can see that the T7 Pro has two more 120-degree wide-angle cameras with 500W pixels than the T7. This is also the biggest difference between the T7 and the T7 Pro. What is the use of it? We will talk about it in detail.
There is a detachable water tank on the “butt” of the Roborock T7 Pro. The constant voltage control water tank separates the water tank and the pallet, and when the pallet is not used, the pallet can be removed to perform general cleaning under the state of water storage. It is particularly worth mentioning that it adopts a drawer design, which can be easily removed without turning over.
Open the top cover, you can see the dust box and the cleaning brush. There is also a Wi-Fi indicator and a reset button at the upper right of the dust box. Dust box and filter can be separated and support washing
This is the situation at the bottom. The steering wheel, driving wheel, main brush, and side brush, the six cliff sensors are unobstructed. The driving wheels can be raised and lowered and rubber tire skins are used, which has a stronger grip and can cross obstacles up to 2cm. Not only can the main brush be quickly taken out, but the bearings on both sides can also be quickly removed, making it easy to clean up the hair.
The Roborock T7 Pro supports sweeping and dragging. The carriage is installed at the rear after being installed. Sweep first and then drag to ensure the cleaning effect. The size of the charging base is larger than in the past, it should be to improve the speed and accuracy of the robot’s recharging.
Three, no longer helpless, choose one to explain in detail the new stone algorithm system that can binocularly avoid obstacles. If you have read the reviews of various types of sweeping robots that the author has written before, I believe you will be impressed with the frequent occurrence of “SLAM”.
SLAM is the abbreviation of “Simultaneous Localization and Mapping”. When translated, it can be called “synchronous positioning and mapping”. It refers specifically to the process of planning a map while the robot calculates its position. Roborock T7 Pro is aimed at where and around the cleaning robot. What is the problem, including the two aspects of positioning and mapping, and based on this planning path to avoid obstacles?
In the past, the solutions that appeared in the sweeping robot can be divided into two traditional factions, one is laser SLAM, and the other is visual SLAM.
The sweeping robots we have tested before are basically all laser SLAM machines. They rely on emitting a laser beam and then bounce back, so as to map the distance between the robot and the target location to form a two-dimensional image. Obtaining the angle and distance of the surrounding obstacles at one time has the advantages of fast acquisition speed and a small amount of calculation, and the disadvantages are also obvious. It is impossible to detect transparent objects and it is difficult to accurately locate in a large area.
Visual SLAM can see the real environment. It uses a camera as a sensor and has a higher degree of discrimination for the environment, but it requires higher computing power. It collects multiple frames of images per second to achieve high-definition camera positioning and synchronizes the SLAM composition. However, a lot of The monocular vision SLAM sweeping robot is difficult to accurately locate in dim or dark environments.
Is there a possibility to combine the video function with the laser SLAM to complement each other’s advantages?
Even Stone himself said that the new RR mason algorithm system is the most critical soft power of Stone. It is responsible for managing and distributing environmental data collected by sensors, and handing over to the “perception and mapping algorithm” for data fusion and analysis, and then “navigation and The “Motion Algorithm” performs dynamic navigation and planning based on the analysis results.
At the same time, it can be noted that the video function of the T7 Pro is not a visual SLAM. The SLAM part is still borne by the laser. After all, the core of visual SLAM is positioning and mapping, but the essence is to use the feature points extracted from the image. The responsibility of the video function of the T7 Pro is to perform AI recognition.
In the test, the author found that under the same size, the T7 Pro’s construction speed and cleaning area are ahead of the past pure LDS lidar products. For example, in some corners, due to table legs and stool legs and other, The interference of debris causes the traditional LDS radar to fail to accurately identify obstacle avoidance and will skip it directly, and the stone sweeping robot can cleverly plan the work path to bypass by virtue of binocular blessing.
Why do binoculars crush monoculars?
The visual recognition system can generally include one or two cameras. A single photo only has two-dimensional information, just like a 2D movie, there is no direct sense of space. We can only rely on our own life experience such as “object occlusion, near and far”. So the information obtained by a single camera and its Limited. And we cannot directly get the effect we want (of course, it can be obtained through some other means, but this item is not yet mature, and there is no large-scale verification).
In analogy to machine vision, the image information of a single camera cannot obtain the distance relationship between each object and the lens in the scene, that is, the lack of a third dimension.
Principle of binocular technology
A single picture is very confusing and uncertain, especially in real-life scenes that are extremely complex. The probability of visual miscalculation from a single camera is very high. And the actual distance of the wrong object may be calculated.
So this requires the robot to be upgraded from a single camera to two cameras, that is, stereo vision, which directly provides information in the third dimension, that is, depth of field information. Therefore, the sweeper equipped with a single camera recognizes all objects as two-dimensional objects, not three-dimensional.
For monoculars, to obtain distance information, you must first identify the target: Monocular ranging and 3D estimation rely on the Bounding Box (BB) of the detection target. Unrecognized objects, the system cannot estimate the distance. For example, if you close your eyes, there is no way to estimate the size and distance of things you do n’t know.
In other words, the monocular sweeping robot can only estimate the pre-trained objects, but the actual home environment is very different, and it is impossible to train all the objects that will cause stuck. It is necessary to assume that the object is on the ground. If it is not, it will cause a ranging error, for example, the object is on a thick carpet. Treating an object as a cube is only a rough estimate, not an accurate distance measurement. It is not known what the object’s actual 3D shape will cause more missed sweeps.
The software algorithm of the Roborock T7 Pro is based on the performance of the sweeper platform. At the same time, it uses the environmental information collected by the vision system (distance avoidance of general obstacles, monocular does not have this feature), combined with AI object recognition algorithm to help the robot accurate Obtained obstacle attribute information, so as to realize active intelligent obstacle avoidance of the sweeper, adopt different obstacle avoidance strategies for different obstacles, evaluate the avoidance distance, and take into account both the sweeping coverage and the obstacle avoidance success rate.
Since AI is used, it means that the sweeping robot can continuously increase the recognition types through the machine learning attribute of AI talent. At present, the T7 Pro has supported weight scales, fan/bar stool bases, shoes, plug-in boards, clews, For 6 types of objects such as 粑粑, OTA plans to add socks, dustpans, U-shaped chairs, and other common debris in the first phase of the follow-up. I believe that more identifiable objects will be added in the future.
Traditional sweeping robots mainly determine obstacles through infrared or ultrasonic scanning sensors, which can identify relatively large obstacles, but they do not know what the obstacles are. This poses a problem, what obstacles need to be avoided, It is impossible to determine what obstacles can be crossed.
The T7 Pro can recognize obstacles like human eyes by using visual recognition technology through the front AI dual camera. Through the above actual measurements, you can see that the T7 Pro perfectly recognizes common obstacles and actively bypasses to avoid Open obstacles.
It is worth mentioning that the T7 Pro adopts different obstacle avoidance strategies for different obstacles and evaluates the avoidance distance to take into account the sweeping coverage and obstacle avoidance success rate: ordinary obstacles < shoes/weight scale/bar stool base / Fan base ＜ Plug-in board ＜ A group of power cords ＜ Baba (the distance between obstacle avoidance is getting bigger and bigger).
It is worth mentioning that in order to better handle binocular data, Stone is equipped with the Snapdragon 8-core processor APQ8053 (that is, the Snapdragon 625 without baseband) for the T7 Pro, and the 8-core Cortex A53 is able to cope with and process image data.
Moreover, T7 Pro can achieve all-weather obstacle avoidance. In order to achieve normal recognition under dim environments such as under the sofa, under the bed, and at night, it uses infrared light to fill the light (infrared is invisible light and will not disturb users).
The dark environment image is still clear and accurately identifies obstacles
According to our understanding, there is generally an IR-cut filter inside the ordinary camera, which brings a contradiction-in order to improve the quality of image generation, infrared light will be filtered out (infrared light is not accepted). In order to receive infrared light in a dim environment.
In order to solve this contradiction. Stone replaced the IR-cut filter with RGB + IR dual-pass filter inside the camera. Under strong light during the day, the infrared fill light does not turn on. And the sensor inside the camera mainly receives visible light. Under dark light, at night The infrared fill light turns on automatically. And the sensor inside the camera mainly receives infrared light. This modification and adjustment of the “big knife” is the only one in the category of sweeping robots.
Daily obstacle avoidance and escape
Under the blessing of the new algorithm led by the binocular obstacle avoidance robot T7 Pro. The stone sweeping robot has more significant improvements in obstacle avoidance and escape. To put it simply, when it encounters obstacles, it will no longer “rump”, but precise strokes. To ensure that the area can be cleaned while passing smoothly without being stuck.
Avoid obstacles while traveling
When it encounters obstacles during cleaning, it can free itself from difficulties. In addition, the T7 Pro also has a certain ability to climb over small steps. It supports up to a 2cm height difference. Some rooms like balcony and kitchen with height differences can also be cleaned freely.
Cleaning and maintenance
Although the sweeping robot is very convenient and fast, cleaning the dust box, filling the water tank. And replacing the mop. These daily cleaning and maintenance require the user to personally complete. And the design is also convenient.
In this water tank design, the design of the Roborock T7 Pro is very intimate. The water tank and the carriage are all designed with drawers. You don’t need to flip the entire sweeping robot when you are installing. And disassembling. Pull the button to pull back, these parts can be easily taken out.
In terms of convenience in cleaning and maintenance, the T7 Pro also performs quite well. The main brush, dust box, mop, these dirty parts can be quickly removed and rinsed with water.
- The dust box can be taken out directly by opening the front cover
- Take the main brush image
- The main brush bearing can be quickly disassembled to facilitate the cleaning of wound hair
In addition, the mop on the carriage is also fixed by the card slot + velcro. After each mopping. You can quickly remove the mop for cleaning. If you find it difficult to clean the mop, you can also choose to use the mop once. You can buy 60 tablets for 99 yuan. Which is cheap and affordable, and enough for you to use for a long time.
In general, the convenient design of the T7 Pro stone sweeping robot in cleaning and maintenance is unique among similar products. This is what a sweeping robot should not be troublesome.
T7 Pro Verdict
The T7 Pro solves the pain points that the sweeper is often stuck and accidentally stepping on the “minefield”. It provides a reliable solution for users who are pursuing cutting-edge technology. And enjoying quality life, especially those who have cute pets and cute babies. s Choice.
For the industry, the T7 Pro adopts AI binocular obstacle avoidance + LDS laser navigation. This solution was previously restricted to a high-priced exclusive cabinet in the industry for a long time. Which is difficult to popularize. This time the stone is responsible for the role of “smashing” and the popularizer-the price is 3899 yuan, and the new product is 3299 yuan. It can be seen that after 2020, more sweeping robot products will follow The footsteps of stone to launch products of related solutions. Leading to a new wave of competition in sweeping robots.
The AI binocular obstacle avoidance adopted by the T7 Pro allows the sweeping robot to really have a pair of “glasses”. At the same time, the addition of infrared cameras allows it to work normally at night and recognize surrounding objects.
When we choose to buy a sweeping robot, motivation is actually very pure. The ultimate goal is nothing more than to improve the quality of life and make ourselves better “lazy.” However, many people fall into misunderstanding when choosing products. Always paying attention to the hardware parameters such as suction and volume. And ignoring the actual user experience of the robot.
Robot “minds” are often more worthy of attention than their hardware performance. In fact, this is precisely the biggest difference between machines from different manufacturers. Nowadays, the hardware performance of sweeping robots is basically the same. At this time, which product allows users to use it most comfortably. The more it will be favored by consumers.
With binocular blessing, the Roborock T7 Pro Vacuum Cleaner is quite fast in both environmental recognition and cleaning speed. The choice of travel path and obstacle avoidance route is also the optimal solution. Every time it runs. It scans the environment in real-time and updates the map For path calculation. You don’t have to worry about it.
Even the most troublesome cleaning and maintenance problem do not exist in the T7 Pro. All key parts of the whole body support quick disassembly. The water tank and the carriage are designed with drawers. They can be removed by pressing and pulling. Both support washing and drying. Even if you are clumsy, these operations are quite easy.
Finally, on the choice of the stone sweeping robot T7 and T7 Pro. We recommend the following: users with less ground debris at home. Users without pets or children recommend the more cost-effective stone sweeping robot T7. If you like cutting-edge technology, there are children or pets, T7 Pro is the best choice. The basic functions of T7 and T7 Pro are the same. And the cleaning and mopping capabilities are the same. T7 Pro only has more AI binocular system and Qualcomm 8-core CPU.
Overall, the Roborock T7 Pro is a bold attempt to sweep the robot product and its form. The new product is only available for 3299 yuan. Which is more expensive than the current high-end flagship sweeping robot market. “True incense” can be said to be just needed for people who like new technology and enjoy a quality life.