Design And Performance Evaluation Of 5G Unmanned Vehicle Test Platform Based On Fog Computing

Unmanned vehicle platforms are widely used in industry,services and special operations fields,and are gradually developing towards autonomous perception,decision-making and multi-vehicle collaborative operation,placing higher demands on the computational processing of unmanned vehicles.In order to reduce the cost and energy consumption of unmanned vehicle platform deployments,offloading of unmanned vehicle computing tasks is critical.However,traditional cloud-based offloading method has poor support for latency-sensitive services and has high data backhaul overhead.To this end,this paper proposes an unmanned vehicle platform design architecture based on fog computing,which offloads the computing-intensive tasks to the fog computing layer.Based on this architecture,an unmanned vehicle test platform based on fog computing is built in both real and simulated environments.In the meantime,relevant functional verification and performance analysis are carried out.The main contributions of this paper include:1.In response to the problems of high information transmission and high overhead of returning data in cloud-based unmanned vehicle platforms,this paper designs and builds an actual unmanned vehicle test platform based on fog computing.In terms of communication,the test platform is connected to the 5G network through a communication module installed on the vehicle.In terms of functions,the test platform realizes the basic functions such as unmanned vehicle motion control,environmental perception,autonomous decision-making and communication interaction,while the computing tasks are flexibly deployed in the vehicle layer,fog computing layer and cloud computing layer according to the delay and calculation amount.Among them,the ultra-low control task is completed by the vehicle layer,the computing-intensive business is offloaded to the fog computing layer,and the massive data processing business is offloaded to the cloud computing layer.The actual measurement results show that for real-time control function module,the delays of control commands from the fog computing layer are typically on the order of tens of milliseconds,while the delays of control commands from the cloud computing layer are typically on the order of seconds.As for the simultaneous localization and mapping(SLAM)function module,compared to being completely processed by the vehicle layer,offloading computing tasks to the fog computing layer can reduce the computing resource consumption of the vehicle layer by 33.5%.2.In response to problems of insufficient automation,low efficiency and insufficient precision of of single unmanned vehicle SLAM,a multi vehicle collaborative SLAM scheme based on fog computing is proposed and the judgement conditions for the end of map building is given.The scheme is verified in the unmanned vehicle simulation test platform.Specifically,this paper designs an autonomous exploration scheme for unmanned vehicles.Each vehicle independently explores the unknown area through the boundary search algorithm.At same time,the vehicle uploads the necessary sensor data to the fog computing layer.After the fog computing layer receives the sensor data,it performs SLAM calculation and local map fusion,and uploads the fusion results to the cloud computing layer.The cloud computing layer receives the local map sent by the fog computing layer,completes the global map fusion and evaluates the global map building effect in point cloud format.The experimental results show that,compared with the single unmanned vehicle SLAM based on pure cloud computing processing,the three unmanned vehicles cooperated SLAM based on fog computing in this paper can save 64.1%of the mapping time,8.9%of the total amount of SLAM service sensor data transmission in average,and greatly reduce the amount of data sent from the fog computing layer to the cloud computing layer.In summary,this paper focuses on the research of an unmanned vehicle platform based on fog computing.In the real environment,the unmanned vehicle platforms is built.Specifically,the unmanned vehicle functional modules is developed,the computing tasks are offloaded to different layers,and performance tests are carried out.In the simulation environment,an automated multi-unmanned vehicle collaborative SLAM scheme is proposed,and the evaluation index of the map building effect is given.This paper provides a certain reference for the application development of unmanned vehicle platform using fog computing.