Using AI to predict changes in wireless communication quality and optimize communicationTechnology to Optimize the Use of Multiple Access Points by Predicting Signal QualityNTT Access Network Service Systems Laboratories

Agricultural and factory IoT networks use wireless communication technologies such as LTE, 5G, local 5G and wireless LAN. In these environments, machine learning technology can be used to predict changes in wireless connection quality, making it possible to switch to a potentially more stable line before the quality deteriorates. Things like video transmission rates can also be controlled according to the predicted values, and the difference between the predictions and actual values can be monitored to detect faults in wireless communications equipment.

In conventional methods for controlling things like autonomous robots and drones, wireless signal quality is observed using moment-to-moment values for factors like signal strength, and devices switch between different types of wireless access and different base stations. However, there are times when the signal quality drops due to an inability to keep up with changes to the wireless environment, and it can be difficult to guarantee the reliability of the connection. In addition, when controlling video transmission rates, it can be difficult to change the video transmission rate quickly enough following a drop in wireless signal quality, even when you dynamically lower the rate following the quality drop. This means it is sometimes not possible to stably transmit video.

• Stable wireless access when controlling drones or autonomous robots using local 5G or wireless LAN
• Anomaly detection in wireless communications equipment based on the difference between predicted values and the actual operational state of wireless access points
• Optimized use of multiple wireless access points for the running of mobile virtual network operators (MVNOs) and overlay VPNs

In the past, decisions concerning which base station a wireless device should use were mostly based on the strength of the signal received from the base station. However, this is no longer a reasonable solution, as the number of wireless-network users has increased dramatically in recent years, meaning that there are now even more conditions to consider, such as the volume of network traffic, the characteristics of different devices and applications, and so on. Our technology leverages machine learning to predict values for wireless signal throughput, latency, jitter and packet loss for each device, and then uses that data to switch to the best wireless base station and select a wireless access method for each application.
The prediction values are calculated by a networked server that functions as a wireless signal quality prediction engine, and which is set up on the cloud or an admin network. Detailed information is periodically captured by the devices and uploaded to the prediction engine. This includes information about location, the type of device and wireless access point, wireless environment scan data, and values relating to actual signal quality. The history information is compiled in a database. The prediction engine processes the data in the database using a machine learning algorithm and produces predictions for the quality (throughput, latency, jitter and packet loss) attainable if the device in question were connected to a base station in range.

NTT Access Network Service Systems Laboratories - Wireless Access Systems Project