Artificial intelligence, especially machine learning, is expected to be actively used in wireless communications. Today's communication systems generate a tremendous amount of traffic data and, in combination with advanced machine learning methods significantly improve the design and management of network and communications components. We are very interested in combining IoT with artificial intelligence technology. Nowdays, the direction of development of IoT is in expansion. There is a saying that the future society will become a hyperconnected society. We have been fascinated by the new useful services created by connecting objects and things, and the value that is gained from Internet of Everything will be beyond them. However, we think that it is important to find out what really useful and important connections are in a society where connection expansion has been realized. Maintaining the myriad connections would be a burden, and new problems, such as the environmental problems that are emerging today, may arise. The connection has been made in pursuit of the usefulness of the connection, but it may be a new burden itself. If so, we may be concerned about disconnecting. I think the artificial intelligence will be at the heart of disconnecting. I think disconnections can be replaced by intelligent connections through artificial intelligence. Let's take a simple example. There is a remote controller on the ground and a drones in the sky. The drones are driven by the remote controller. The drone waits for a command from the remote controller to take a special action. But what if the wireless channel is unstable and the connection is disconnected from time to time? The drones are waiting for the command from the remote controller for the next operation, but the command is not received. The drones will be able to use the history of the current situation and the past command to estimate the command to be entered in the future. The better the drone can infer, the better we can do what we want to do, even if we are connected to an unstable wireless channel. In this case, we can go one step further and make some assumptions about the commands we want to send from the remote control. If it is determined that there is a high possibility of estimation, it may not be transmitted intentionally. It is also used not to use unstable channels, it can reduce the energy consumed in communication by not using communication resources (frequencies), we can reduce radio interference and yield to other users / applications as needed. It is very important to reduce the burden of communication if the drones are not one, but more than thousands. It's like telepathy.
The ultra-reliable communication functionality is a key enabling factor for the wireless networked control systems, which is one of the variants of IoT. For example, intelligent transport system, exploration of disaster/pollution area, and safety and surveillance system, a number of mobile vehicles and flying drones can be used to discover the large area. The command from the control tower or other machines and sensing data are delivered via wireless link and these messages should be delivered with very high reliability. Due to the mobility of machines, the network will hardly maintain. The transmitted messages are at a very low data rate and short packet length, but there will be a lot of message exchanges. It could be severe burden sending the meta-data packets for the purpose of connection setup (e.g., preambles, time sync frames, etc.). To alleviate the situation, the new communication design is needed. At the MAC layer, nodes can transmit randomly, without any prior scheduling strategy, which saves resources needed for coordination between communication nodes. The research for the new kinds of transmission strategies like joint data/meta-data encoding and new coding techniques for short blocklength would be needed. Consequently, transmitted packets might encounter collisions, which is the first capacity and reliability limiting factor in this kind of network. The advanced receiver design utilizing signal processing and successive interference cancellation is highly desirable. The goal of this research is to characterize and improve the network performance for the applications demanding ultra-reliable and/or low-latency communication in the future wireless IoT systems.
Compared to the previous generations of wireless systems, 5G has an increased focus on Internet-of-Things (IoT) connectivity. Specically, two of the three generic 5G services, URLLC and massive MTC (mMTC) are related to IoT and those are our main research topics. Traditionally, the design and optimization of communication systems are based on statistical models, which becomes highly suboptimal with the increase of the complexity of the networks, heterogeneous services and massive number of connections. Therefore it becomes necessary to base the optimization on data analysis and machine learning techniques. we will work with data-oriented models to optimize the operation of wireless IoT systems. It will be focused on, but not limited to the emerging 5G systems. As example, dataoriented approaches can be applied to optimize the exchange of control information in protocols or integrate the communication protocol with the process of mining of the data stream that comes from the IoT devices.
TBD