Nearmap partnership to accelerate 5G with digital twins

Nearmap and Digital Twin Sims have partnered to create accurate digital twins of U.S. cities to help telcos roll out 5G and new IoT services. Nearmap’s automation pipeline transforms high-resolution imagery and spatial data into 3D models for city planning, construction and urban planning. Digital Twin Sims is a telecommunication business and network modeling advisory firm. 

Digital Twin Sim’s cofounder, Sameer Lalwani, started the company after a 25-year career in telecom planning. The industry has long used manual approaches to craft models that started at 100 meters of resolution and gradually improved to 10 meters. Over the last few years, he saw an opportunity to take advantage of high-resolution data from companies like Nearmap as well as cloud computing from the likes of AWS and Microsoft Azure to scale improve this resolution down to 15 cm at city-scale. 

Planning for complexity

5G deployments create a new level of complexity and demands greater precision than traditional mobile networks. For starters, 5G requires far more base stations to reach the same levels of coverage as previous wireless services. Additionally, many spectrum bands are more susceptible to interference or reflection by buildings or even tree branches. As a result, telcos need to ensure line-of-site paths between cell towers and coverage zones — something that has traditionally been an expensive and time-consuming process

Nearmap’s platform automatically feeds data into the Digital Twin Sims engine to generate models that reflect changes, including new construction and even the growth of trees and other vegetation that might impact coverage. The platform can constantly update these models in response to changes detected in aerial and satellite surveys. This new approach allows Digital Twin Sims to simulate thousands of nodes in a single afternoon, whereas  a manual survey would traditionally require several days to plot a single node in a 5G network.

The digital twins combine physical models of building and infrastructure — down to the lamp posts — with demographics and business models representing existing customers and insights into new opportunities. Better simulations will help telco executives evaluate business opportunities and will allow engineering teams to plan and execute optimized infrastructure. 

The digital twin creation process starts with telco planning teams setting a particular goal. A strategic overview for a nationwide network starts with low-resolution data for high-level planning. A more tactical analysis for estimating the cost and time to provision a millimeter-wave or 5G network would begin with the highest-resolution possible. 

Currently, operators, marketing, sales, network planning, deployments and device purchasers operate in different business process and data silos. Each group is working towards different metrics that, at times, conflict with other departments. 

“There needs to be a single entity where the datasets from all these sources are combined and looked at holistically, but that rarely happens,” Lalwani told VentureBeat.

Different formats required

Digital Twin Sims creates a consolidated view of data from many sources and various formats presented within a single UI. For raster data, such as satellite and aerial imagery, they use GeoTIFF and are now moving to Cloud Optimized GeoTIFF (COG). All vector data representing the routes of lines, properties of buildings and demographic data, is stored in GeoPackage (GPKG) files. Point cloud data derived from lidar scans is also captured and stored in .laz files. 

The digital twin downloads data from the Nearmap API and loads it into an H3 server at a fixed resolution. H3 is an open source geospatial indexing system developed by Uber. Lalwani said they usually customize their existing code for a specific scenario and run it in Docker containers based on the size of computing needed.

Nearmap’s general manager of North America, Tony Agresta, said the company prioritizes having a robust distribution of file formats that can easily be integrated with as many third-party applications as possible. They have developed tools to translate data feeds into standard ortho imagery (as viewed from directly above), 3D data created through photogrammetry and vector AI data. 3D formats include textured mesh OBJ, SLPK, 3MX, Cesium and FBX files and point cloud data in a LAS file. 3D digital elevation models, digital terrain models (bare earth) and true ortho are all stored as GeoTIFF. Vector AI is stored in GPKG and Shapefiles.

As it is still relatively early in mass use and adoption of digital twins, best practices for developing this technology are still a work in progress. Engineers, data scientists and executives need to keep their options open for adopting the best data formats required to simplify data pipelines, create useful simulations and provide the greatest insight for each use case. Down the road, emerging efforts like universal scene description (USD) could make it easier to transform data across use cases.