A multi-vehicle collaborative mapping method for autonomous driving

Abstract

The present invention relates to the technical field of intelligent driving, and more specifically, relates to a multi-vehicle cooperative mapping method applied to automatic driving. Including the following steps, step 1: data acquisition and perception; step 2: preprocessing of point cloud data; step 3: local map and global map; step 4: communication module; step 5: matching between vehicles; step 6: matching After success, each vehicle sends the matching result to the other party; Step 7: After each vehicle receives the global edge map and the other party's trajectory data that are continuously sent by the other party, it converts the data through the matrix calculated in step 5, and then converts the result It is transmitted to the mileage calculation and mapping part for real-time collaborative mapping.

Description

technical field [0001] The present invention relates to the technical field of intelligent driving, and more specifically, relates to a multi-vehicle cooperative mapping method applied to automatic driving. Background technique [0002] LCM is a set of lightweight libraries and tools for messaging and data marshalling that can target high-bandwidth and low-latency real-time systems. It provides a publish / subscribe messaging model and automatic marshalling. LCM is implemented through UDP The multicast mechanism transmits customized data information under a certain multicast address and channel. Therefore, this method takes advantage of its real-time and multicast characteristics, and chooses it as a method for cooperative communication between vehicles. [0003] ROS (Robot Operating System) is a robot operating system. It provides the services that an operating system should have, including hardware abstraction, low-level device control, implementation of common functions, ...

AI Technical Summary

Problems solved by technology

[0006] The main problems of some current technologies are as follows: most of the current mapping technologies are aimed at the construction of bicycles; the mapping scheme needs more peripheral sensors, which are expensive; The problem of slowness and low efficiency; the existing collaborative mapping methods require a large amount of memory space, a large amount of calculation, and high CPU requirements; the existing collaborative mapping methods need to occupy a large amount of bandwidth resources;

Although this approach can maximize the detection field of view of a vehicle, the use of multiple sensors in a single vehicle has limits and limitations, and it is impossible to make further breakthroughs in the detection range.

[0008] In the process of building a single-vehicle map, with the prolongation of the cumulative running time of the algorithm, the cumulative error of the algorithm itself is getting larger and larger, and the accuracy of the built map is reduced, and due to the long-term single-vehicle mapping, it is easy to cause system instability

[0009] The existing multi-vehicle collaborative map construction method is based on graph features. The algorithm requires vehicles to send their own graph nodes to the entire network anytime and anywhere during the entire process, resulting in a decrease in network utilization and tight network bandwidth resources.

At the same time, after all the nodes in the network receive the data sent by the other party, they directly perform a large number of complex calculations without any preprocessing and other optimization operations, which requires a very high CPU cost for running the algorithm.

Images