The mobility of unmanned aerial vehicle (UAV) guided service delivery, combined with 5G communication networks, provides new methods to supply real-time services for disaster management systems and e-healthcare monitoring.
But there are challenges to fully meeting these benefits. The networks that serve UAVs face issues of low service and delays due to limited resources. The urgent nature of emergency situations and the critical needs of health care require intelligent resource management that prioritizes needs. Dr. Uttam Ghosh, associate professor of cybersecurity, and colleagues used machine learning techniques to develop the solution, SoftDrone: Softwarized 5G Assisted Drone Networks for Dynamic Resource Sharing.
A programmable network, with dedicated network slices for UAV-based, on demand communication layers, will improve overall network performance.
“However,” says Ghosh, “increasing random service demands, network resource allocation, retention, and release are serious networking challenges. Existing methods often consider dedicated resource allocations and that causes poor resource utilization and service quality.”
Machine learning techniques have been used to improve performance. But, this method has hurdles to overcome. “Limited energy constraints and complexity in resource cycle management are a concern,” says Dr. Ghosh.”
Dr. Ghosh and colleagues Deborsi Basu, Soumyadeep Kal, and Professor Raja Datta from the Indian Institute of Technology in Kharagpur, India, have proposed an intelligent, data-sharing scheme to address these network service issues.
“During time-critical situations and disaster management, the selection of the service requests becomes extremely crucial. We must give priority to the resources responsible for those services, ” says Dr. Ghosh.
Dr. Ghosh and his colleagues pursued a machine learning algorithm that improves the prediction of resource requirements, according to network traffic load, with fewer errors. They found that the regression model trained the data much quicker than previously proposed algorithms in this area.
“The advantage of training the model faster,” says Dr. Ghosh, “combined with a much more accurate resource requirement prediction, according to network traffic load, provides a more robust solution than previous work in this area.”
They also developed the network in a modular way so it can be used in tandem with other possible types of scaling and placement algorithms. This approach will provide a more complete solution to resource allocation and placement.
“Developing countries need rapid recovery with a smart state of the task force. By introducing this type of technology, we can reduce the casualty rate to a huge extent,” says Dr. Ghosh.
The full paper, SoftDrone: Softwarized 5G assisted drone networks for dynamic resource sharing using machine learning techniques,is available online.