Wireless Innovation Forum Top Ten Most Wanted Innovations

Innovation #3: Propagation Prediction Techniques for Dynamic Spectrum Sharing

3.1 Executive Summary

To enable next-generation, high-capacity mobile wireless networks to dynamically share radio spectrum with incumbent users, faster and more accurate propagation prediction techniques are needed to perform real-time interference calculations in support of automated frequency assignment. These networks consist of many moving and spectrally and/or directionally agile nodes, which means that propagation conditions can be subject to rapid variations; typically, these nodes are also concentrated in urban environments. Due to the high complexity of such environments, currently available prediction techniques are either too slow (for example, ray tracing models) or too inaccurate (for example, empirical path loss formulas) to support efficient spectrum sharing.

3.2 Application

The application of this innovation is in database-driven dynamic access to shared spectrum by the nodes of next-generation commercial mobile wireless networks deployed in urban areas. One possible sharing scenario, for example, is an urban 5G deployment in a frequency band whose incumbent users are fixed and located in the surrounding rural area. Spectrum resource assignments to the mobile nodes are controlled by an automated, database-driven decision-making entity protecting the incumbent users from harmful interference. When a mobile network node requiring access to spectrum resources registers or updates its operational parameters (location, transmit power, antenna parameters, etc.), this entity must be able to perform nearly instantaneous calculations of the resulting interference impact on the incumbent users. These calculations must take into account the interference shielding effects of the urban environment, which are due to signal obstruction by obstacles such as buildings. Improved propagation models to more accurately characterize radio signals in small-cell deployments in dense urban environments would benefit service providers to aid them in more effective network deployment which ultimately would benefit consumers (both business and the general public) with improved coverage and quality of service. It would also give the regulators a better understanding of radio propagation characteristics in urban environments that would aid them in their decisions on how best to allocate spectrum so as to use the spectrum most efficiently in these environments.

3.3 Description

For propagation prediction methods to be of value for database-driven dynamic spectrum sharing among mobile radio nodes, they need to be:

  1. Sufficiently fast to support spectrum assignments based on the actual locations of mobile nodes; for vehicular nodes, this implies prediction times of less than one second.
  2. Sufficiently accurate to prevent harmful interference while not being overly conservative with respect to estimating propagation loss on interference paths, which limits spectral utilization efficiency; prediction uncertainty should be dealt with statistically rather than by applying safety margins.

These requirements are difficult, if not impossible, to meet using conventional prediction methods suitable for urban environments, because deterministic prediction of the multipath structure in such complex environments is computationally very expensive. Possible R&D directions for overcoming the limitations of traditional prediction techniques include the use of:

  • High-performance cloud computing to accelerate existing prediction methods.
  • Radio-frequency sensing data collected from mobile devices in the frequency sharing area, in the same or different frequency bands, supplemented by deterministic-based simulations to determine probability of achieving coverage extent, quality of service indicators etc,
  • Maps of buildings, terrain and vegetation, as well as other sources of external data; for example, dynamic maps of vehicular traffic intensity (possibly relevant for signal obstruction) and weather (possibly relevant for millimeter-wave propagation).
  • Machine learning techniques for recognizing patterns in past propagation prediction results and their correlation with external data.