A Semantic Map Construction Method Based on Cloud Robot Hybrid Cloud Architecture

Abstract

The invention discloses a semantic map construction method based on cloud robot hybrid cloud architecture, aiming to achieve a proper balance between improving the accuracy of object recognition and shortening the recognition time. The technical solution is to build a hybrid cloud composed of robots, private cloud nodes, and public cloud nodes. The private cloud nodes obtain environmental pictures, odometer and location data taken by the robot based on the ROS message mechanism, and use SLAM to draw the environment in real time based on the odometer and location data. Geometric map. The private cloud node performs object recognition based on the environment picture, and uploads the object that may be misrecognized to the public cloud node for recognition. The private cloud node maps the object category identification label returned by the public cloud node with the SLAM map, and marks the object category identification label on the corresponding position of the map to complete the construction of the semantic map. By adopting the invention, the local calculation load of the robot can be reduced, the request response time can be minimized, and the accuracy of object recognition can be improved.

Description

technical field [0001] The invention relates to the technical field of robot distributed computing, in particular to a method for building a cloud-based robot semantic map by using cloud computing as backup support and building a hybrid cloud architecture. Background technique [0002] The perception data sources of robots may include multiple dimensions such as vision, force, touch, infrared, ultrasound, and lidar. Robot semantic map construction means that robots recognize and understand the environment based on these perception data, and the core focus is on how to perceive The data is extracted from the rough and the essence, the false and the true are removed, and then analyzed and synthesized to extract high-level semantic information (such as object names and locations) that can be used by the robot for autonomous decision-making. The specific performance is to add labels of objects in the environment on the geometric map. The acquisition of semantic information can b...

AI Technical Summary

Problems solved by technology

[0006] To sum up, the solution based on private cloud performs better in request response time, and can recognize familiar objects, but requires pre-training, and has knowledge limitations, and cannot recognize unfamiliar objects in an open environment; the solution based on public cloud utilizes Internet big data knowledge has broader intelligence for object recognition in the environment, and can recognize unfamiliar objects without pre-training, but the recognition delay is large and the request response time is long

Images