Shulou(Shulou.com)06/02 Report--

Positioning technology is the most basic link for the robot to realize autonomous positioning and navigation, which is the position and posture of the robot relative to the global coordinates in the two-dimensional working environment. At present, SLAM (Simultaneous Localization and Mapping real-time positioning and map construction) is the mainstream positioning technology in the industry, which can be divided into laser SLAM and visual SLAM.

What is laser SLAM?

Laser SLAM originated from the early ranging-based positioning methods (such as ultrasonic and infrared single-point ranging). With the emergence and popularization of lidar (Light Detection And Ranging), the measurement is faster and more accurate and the information is richer. The object information collected by lidar presents a series of scattered points with accurate angle and distance information, which is called point cloud. Usually, the laser SLAM system completes the positioning of the robot itself by matching and comparing two point clouds at different times and calculating the distance and attitude change of the relative motion of the lidar.

The lidar distance measurement is more accurate, the error model is simple, the operation is stable in the environment other than the strong light, and the point cloud processing is relatively easy. At the same time, the point cloud information itself contains direct geometric relations, which makes the path planning and navigation of the robot intuitive. The theoretical research of laser SLAM is also relatively mature, and the landing products are more abundant.

What is visual SLAM?

The eye is the main source for human beings to obtain external information, and visual SLAM also has similar characteristics. It can obtain massive and redundant texture information from the environment and has a strong ability to identify the scene. The early visual SLAM is based on filtering theory, and its nonlinear error model and huge amount of computation have become obstacles to its practical landing. In recent years, with the sparse nonlinear optimization theory (Bundle Adjustment) and the progress of camera technology and computing performance, real-time running visual SLAM is no longer a dream.

Usually, a visual SLAM system consists of a front end and a back end. The front end is responsible for calculating the position and pose of the robot through visual increment, which is fast. The backend is mainly responsible for two functions:

First, when there is a loop (that is, it is determined that the robot has returned to the vicinity of the previously visited location), it finds the loop and modifies the position and posture in the middle of the two visits.

Second, when the current tracking is lost, the robot is relocated according to the visual texture information. In short, the front end is responsible for rapid positioning, and the back end is responsible for slower map maintenance.

The advantage of visual SLAM is the rich texture information it uses. For example, two billboards with the same size but different content can not be distinguished by the laser SLAM algorithm based on point cloud, but can be easily distinguished by vision. This brings unparalleled advantages in repositioning and scene classification. At the same time, visual information can be easily used to track and predict dynamic targets in the scene, such as pedestrians, vehicles, etc., which is very important for the application in complex dynamic scenes. Third, the visual projection model theoretically allows infinitely distant objects to enter the visual picture, and large-scale scene positioning and map construction can be carried out under reasonable configuration (such as long baseline binocular camera).

For a long time, the industry on laser SLAM and visual SLAM in the end who is better, who is the mainstream trend in the future have their own views, the following will be a simple comparison from several aspects.

Application scenario

In terms of application scene, the application scene of visual SLAM is much richer. Visual SLAM can work in both indoor and outdoor environments, but it is highly dependent on light and can not work in dark places or some untextured areas. At present, laser SLAM is mainly used indoors for map construction and navigation.

Accuracy of positioning and map construction

In static and simple environment, laser SLAM positioning is generally better than visual SLAM, but in large-scale and dynamic environment, visual SLAM shows better effect because of its texture information. In the map construction, the accuracy of laser SLAM is higher, and the accuracy of the map constructed by RPLIDAR series of domestic Silan technology can reach about 2cm. And the visual SLAM, such as the common depth camera Kinect, (ranging range is between 3-12m), the accuracy of map construction is about 3cm; therefore, the accuracy of the map constructed by laser SLAM is generally higher than that of visual SLAM, and can be directly used for positioning and navigation.

Ease of use

Both laser SLAM and visual SLAM based on depth camera directly obtain the point cloud data in the environment, and measure where there are obstacles and the distance of obstacles according to the generated point cloud data. However, the visual SLAM scheme based on monocular, binocular and fisheye cameras can not directly obtain point clouds in the environment, but form gray or color images, which need to constantly move their own position, extract and match feature points, and use triangulation to measure the distance of obstacles.

In addition to the above points, there is also a certain gap between laser SLAM and visual SLAM in terms of detection range, operation intensity, real-time data generation, map cumulative error and so on.

Note: left is Lidar SLAM, right is visual SLAM, data source: KITTI

It is obvious that for the same scene, there is a deviation in the second half of the visual SLAM, which is caused by the cumulative error, so the visual SLAM needs loop test.

Generally speaking, laser SLAM is a relatively mature robot positioning and navigation technology, and visual SLAM is the mainstream research direction in the future. In the future, multi-sensor fusion is an inevitable trend. Learn from each other's strengths and combine their advantages to create a really easy-to-use and easy-to-use SLAM solution for the market.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.