RockMass is working to break into the niche market of mining and tunnel engineering.

This Toronto-based startup is leveraging NVIDIA AI to develop a mapping platform that helps engineers assess the stability of mine shafts and tunnels under construction.

Currently, as a safety precaution, geologists and engineers stand five meters away from the rock to visually assess the hazard level of rock formations. However, Shelby Yee, co-founder and CEO of RockMass, believes this is not an ideal way to ensure the accuracy of results.

"Using their current assessment method, the entire process takes nearly 90 minutes, whereas our technology can complete it in about 5 minutes," Yee said.

RockMass is testing its handheld device, Mapper, with field engineers. The device is applicable in mining, geological exploration, and civil engineering. The startup is developing an AI platform for robots, drones, and handheld devices that capture geological data.

Now, the startup's Mapper AI device offers a safer way for engineers to stay away from potentially collapsing tunnels, while also providing a faster data collection and processing system. Robots and drones using this platform can enter higher-risk areas.

RockMass's clients include Brazilian mining company Nexa Resources, which aims to leverage RockMass's technology to enhance its automation and safety levels.

AI for Geotechnical Engineering

For years, engineers have used traditional equipment to measure the angles of rock surfaces, such as optical survey equipment mounted on tripods, similar to a theodolite. They need to identify so-called weak planes to determine fracture points within tunnels and rock formations.

Engineers measure rock surfaces and collect data to construct what is known as a stereonet. A stereonet can represent three-dimensional shapes (such as boulders) on a two-dimensional display.

Matt Gubasta (Co-founder and CFO) tests the instrument at an underground testing center in Sudbury, Ontario.

Traditionally, engineers had to bring data collected from the field back to the office and transfer it to a computer to construct a stereonet.

The startup's technology offers a simpler method. Its handheld device is equipped with sensors that can perform such measurements. Its LiDAR sensors and inertial measurement units (IMUs) can map the orientation of weak planes in rock formations. Furthermore, the device can operate normally even in underground environments without GPS, wireless communication, or light.

By utilizing the information provided by these sensors, RockMass's software can quickly identify usable data for engineers within minutes. The company is dedicated to helping field engineers capture and process data on-site. "We can view the data in real-time," Yee said.

Computationally Intensive AI

According to Stuart Bourne, co-founder and CTO, RockMass's platform for collecting field data is very demanding in terms of computational power. The company's devices rely on the robotic performance of NVIDIA Jetson, powered by CUDA, cuDNN, and TensorRT software libraries.

"Jetson's computational power is extremely high relative to the energy it consumes," Bourne stated.

The startup utilizes the CUDA library to process data in real-time within cloud instances running NVIDIA GPUs, thereby processing stereonets for clients.

"No one else collects and processes data like we do," Yee said. "We can process data in the cloud in real-time, thanks to the computational power of GPUs."