The emergence of smart home products, so that "lazy cancer patients" have seen the gospel, especially intelligent cleaning products in the star product, sweeping robot, almost become a necessity of life. However, due to the lack of technology accumulation in the early stage, the market has risen prematurely and capital has poured in, resulting in a chaotic situation where "technology is not enough to make up for the concept", leading to a sharp decline in market reputation.

At present, the core competitiveness of the sweeping robot market is mainly reflected in the cleaning ability, navigation and obstacle avoidance, intelligent interaction, with the sweeping and dragging of one machine with automatic dust collection function, cleaning ability is no longer a problem, but navigation and obstacle avoidance is still a market pain point, intelligent interaction is still at a relatively initial stage.

In response to the above industry pain points, INDEMIND has developed a binocular vision navigation solution for sweeping robots, which provides functions from map building to path planning, intelligent identification, intelligent obstacle avoidance and interaction through stereo vision and artificial intelligence technology.

01

3D stereo vision for global planning and cleaning

INDEMIND binocular vision navigation uses unique stereo vision and VSLAM algorithms to solve 2D images of the environment taken by the left and right eye into 3D spatial structures with depth information, making it more suitable for sweeping robots to achieve better accuracy and stability of navigation and positioning, and to accurately cover the whole house.

The INDEMIND binocular vision solution supports 3D stereo mapping
After processing, it outputs a 2D raster occupancy map and can provide a variety of path planning modes according to the needs of the scene and cleaning mode, for path planning of the floor sweeper, etc. It also supports multiple map saving, repositioning and intelligent no-zone functions.

In the global sweeping process, the sweeper adopts a bow-shaped sweeping method, with a clear sweeping path, high operational efficiency and wide coverage.

02

Binocular stereo vision for more accurate obstacle avoidance

The INDEMIND binocular vision navigation solution fundamentally improves the sweeper's ability to perceive the environment, providing a depth resolution function of <1% in the range of 0.05-1.5m, together with a horizontal 60° and vertical 30° FOV, which can cover areas that cannot be covered by traditional sweeper obstacle avoidance solutions, providing full range obstacle detection for the sweeper.

INDEMIND has developed a lightweight object recognition convolutional neural network model for home scenes, which can quickly and accurately identify objects, people, animals and environmental semantics after detecting obstacles, with recognition accuracy ≤ 2cm, and can identify more than a dozen major categories and hundreds of common household objects, avoiding sweeping robots from dragging obstacles or getting stuck. It is also able to strategically avoid specific obstacles such as pet droppings, fragile objects and wires at different obstacle avoidance distances.

Tested in a real home environment, the INDEMIND binocular vision navigation solution was able to accurately avoid every obstacle during a complete cleaning process, with no false touches or collisions, making it an excellent obstacle avoidance solution. At the same time, in the face of dense and complex environments, the sweeper relies on its unique VSLAM algorithm, which is able to operate stably and achieve fast and accurate obstacle avoidance when dealing with large volumes of highly complex environmental information.

In addition, as the vision sensor is susceptible to changes in light, the INDEMIND binocular vision module is equipped with an adaptive infrared fill light, which enables accurate identification of obstacle avoidance in both bright light and dark and weak textured environments.

03

Intelligent upgrade and interaction first

With the binocular camera, the robot can also be remotely controlled by an app, so that it can observe its surroundings in real time during the cleaning process. At the same time, it can also achieve remote care, intelligent security and other extended interaction, to observe the home environment at any time and anywhere.

In addition, it can also be controlled by voice and other commands to achieve specific cleaning, such as intelligent following, designated area cleaning, etc., so that you can do whatever you want without having to do anything.

Overall, unlike other navigation solutions, the binocular vision solution is more functional, more expandable and less costly, making it the best all-round navigation solution for floor cleaning robots.