Laser SLAM map method based on multi-robot cooperation

Abstract

The invention discloses a laser SLAM map method based on multi-robot collaboration. The method comprises the steps of achieving the grid conversion of an SLAM map; representing the grid map splicing problem by using a minimization problem of image registration, establishing a mathematical model, and providing an iterative solution method based on an ICP algorithm; extracting local invariant feature points from multiple targets in the grid map to be spliced by utilizing a map matching algorithm based on a convolutional neural network, establishing a vector set from template feature points to the center of the template, and estimating the central position of a sub-graph copy by combining and searching the corresponding matching point position in the global map, and taking the central position as an initial splicing parameter; taking the splicing parameter as an initial value of an ICP algorithm, solving an objective function, and finally realizing the fusion from the grid sub-map to thegrid global map; realizing the compression and extraction of the map by utilizing the sparse representation of the map and a low-rank matrix reconstruction algorithm. The method is higher in adaptability, is easy to process and is low in cost, and the storage capacity is reduced.

Description

technical field [0001] The invention belongs to the technical field of laser SLAM maps, in particular to a laser SLAM map method based on multi-robot collaboration. Background technique [0002] In recent years, with the steady advancement of the application of SLAM on single robots, the concept of applying existing SLAM methods to multi-robot systems has attracted more and more attention. Compared with single-robot systems, multi-robot systems have advantages in terms of execution efficiency, fault tolerance, robustness, reconfigurability, and hardware costs, and can perform emergency rescue, disaster relief, and resource management in unknown environments that are difficult for humans to reach. However, it is still difficult to realize multi-robot collaborative work in an actual large-scale environment. The main reasons can be summarized as follows: [0003] (1) Cost issue. The cost of a single robot with SLAM (including hardware, software and intellectual property right...

AI Technical Summary

Problems solved by technology

The cost of a single robot with SLAM (including hardware, software and intellectual property rights) is too high, especially the market price of a robot with 3DSLAM is more than one million, so it brings certain resistance to the promotion and industrialization of multi-robot systems

[0004] (2) Communication problems

In the actual large-scale environment, the communication range between robots is often limited, or when there is a lack of geographic information (such as GPS or Beidou positioning information cannot be obtained indoors), all SLAM algorithms that assume unlimited communication between robots will lose their meaning

[0005] (3) Data association problem

The usual SLAM algorithm is to avoid the defect of data association problem through the observation value of the sensor and the preset landmark of the environment, but when the preset landmark does not exist in the unknown environment, or there is no prior knowledge of the relative positions between multiple robots Under such circumstances, how can the robot locate itself on the map without the so-called "missing" problem?

[0006] (4) Map storage problem

[0007] (5) Map fusion problem

Due to the randomness of robot placement, the variability of the size of the environment, and the scalability of multi-robots will bring resistance to the solution of the map information fusion problem, and the current research on this problem is still being explored, so the map information fusion The problem has become a bottleneck in the development of mobile robots, especially multi-mobile robot SLAM applications

Images