But sometimes you might encounter a robot that suddenly stops walking or makes a sudden turn, failing to navigate correctly to your destination. Are all robots these days so "bad"? Actually, what bad intentions could a robot possibly have? It just needs a "perfect" map.
The quality of a map directly determines a robot's subsequent localization and navigation capabilities. So how do we create a "perfect" map?
▲Collection of SLAMTEC Robotics Mapping
#01
No matter what kind of map it is, first and foremost, it must be a complete map.
However, many large-scale scenarios, such as large shopping malls and large stadiums, are too large and require higher mapping performance from intelligent robots. The first step for excellent intelligent robots to take on this challenge is mapping of "large-scale scenarios".
SLAMTEC's robots can map areas up to 500m x 500m and are operating stably in two large makeshift hospitals in Lingang and Qingpu. Building large-scale maps presents many challenges, with the lack of a closed-loop mapping system being one of the most significant and persistent technical difficulties.
So how do we teach robots to "actively close loops"?
▲ SLAMTEC Robotic Mapping Active Closed-Loop Demonstration
SLAMTEC's mobile robot chassis is equipped with the SLAM 3.0 intelligent navigation algorithm, enabling active loop closure detection. Upon detecting new loop closure information, it can effectively correct the post-loop closure map, achieving more reliable environmental mapping.
Here are some mapping tips to help you create a more complete and closed-loop map.
1►
First small closed loop, then large closed loop
Try to control the robot's movement to form small, closed loops. After completing the closure of a small loop, gradually expand outwards to build a map. Avoid directly attempting to close larger loops, as excessive accumulated errors will directly lead to loop closure failure.
▲Large cumulative error leads to closed-loop failure
Avoid taking paths unrelated to the current loop closure, as this will generate accumulated errors. This can easily lead to a large discrepancy between the beginning and end of the loop, making it impossible to close the loop.
2►
Select points with rich features as the closed-loop points.
When creating a map and closing the loop, try to select areas with abundant laser points as the loop closure points, and avoid selecting points with similar features such as long straight corridors as loop closure points, as such environments are prone to incorrect loop closure.
▲Complete the mapping loop on the side that is closest to the environment.
When the surrounding environment features are weak, try to place the robot closer to the most prominent side of the environment to complete the mapping. During mapping, try to move the robot in a straight line to avoid sparse feature points caused by the robot's rotation.
3►
Explore overlapping paths and refine the details
After the loop returns to the origin, keep the robot moving, taking overlapping paths as much as possible. Do not stop moving immediately. On the already closed path, further scan the map to refine the details.
▲Further refine map details
※ Avoid circling or backtracking before the loop is closed, as this can easily lead to loop closure failure or errors if there are few environmental features available.
In shopping mall environments with a lot of glass, if there are distinctive features ahead of the robot, such as pillars, left and right passages, or slanted walls, it can stop at a suitable location, rotate in place towards the feature, wait for the radar to scan the feature, and then rotate back to continue moving forward. However, try not to move backward during the rotation.
#02
Therefore, radar scanning to identify environmental features is also crucial. Building upon the closed-loop mapping of large scenes, SLAMTEC, through years of technological upgrades and leveraging LiDAR sensors, has achieved even greater precision in autonomous mapping for its robots.
1►
Precise detection of small objects
Even tiny objects that are easily missed by the naked eye can be accurately detected by SLAMTEC's lidar.
The Slamtec chassis is equipped with the Slamtec S2 self-developed LiDAR, which has a sampling frequency of up to 32 kHz and can collect 32,000 ranging data per second, so even the smallest object cannot escape its detection.
▲ S2. Small object detection effect
Meanwhile, it is equipped with two depth camera systems, perfectly combining LiDAR and multiple visual sensors, enabling 360° all-round navigation and obstacle avoidance, allowing the robot to map and navigate more stably and accurately.
2►
Precise detection of black substances
Because robots rely on LiDAR to emit light to acquire data for mapping, but black objects easily absorb light, leading to detection errors. However, there are many black objects in real life, such as black handrails and black seats. If the robot cannot detect these black objects, it will collide with them while walking.
▲ S2 measures black and white objects without layering
The Slamtec mobile chassis is equipped with a self-developed LiDAR, the S2, which can effectively detect black objects. At the same time, it can also accurately detect highly reflective objects, such as common glass and mirrors, which are unavoidable in various scenarios, helping the robot achieve more accurate positioning, navigation, and obstacle avoidance.
The above indicators and data demonstrate that SLAMTEC's mobile robot chassis is more than capable of assisting robots in autonomous and detailed mapping. Intelligent service robots developed based on SLAMTEC's robot chassis can possess a "perfect" map, enabling them to perform autonomous positioning and navigation more effectively. Therefore, the "malicious intentions" you may have encountered with robots in the past will no longer exist.
