Emesent is a drone autonomy, mapping, and analytics company—and an Energy Drone & Robotics Coalition partner. Here’s what we learned in an interview with Stefan Hrabar, CEO of Emesent.

Don’t want to miss future topics like this? Check out our upcoming events. 

Hrabar and his co-founder, Farid Kendoul, founded Emesent to commercialize technology they had worked on for over 10 years at Australia’s Commonwealth Scientific and Industrial Research Organisation (CSIRO). Thus was born Hovermap, a smart mobile scanning unit which mounts on drones to autonomously capture insights in challenging and hard-to-reach industrial environments.

State of Drone Autonomy

Drone autonomy can be compared to the levels for autonomous vehicles, where level 0 is fully reliant on a human operator and level 5 is fully autonomous in any environment or situation. 

A level 0 drone would be one that is fully controlled by a pilot at all times. Level 5 drone autonomy is possible, but Hrabar pointed out, “Today we are seeing systems in the upper levels of autonomy around level three and level four, including what we're doing at Emesent, where a pilot is not needed for an entire mission, but those are for very specific mission scenarios. It’s not like the system can fly in any environment at any time.” 

And those mission-specific use cases are where companies such as Emesent are investing heavily in industrial drone autonomy, as this will offer significant value to customers.

Mission Appropriate Autonomy

Full drone autonomy can offer many benefits, but it isn’t always the best solution. If a system is running without the possibility for human intervention, the mission can’t be adapted to meet unforeseen requirements. Hrabar said, “Relying on full autonomy can be It ends up being less efficient because if your requirements change throughout the mission, you then have to re-send the system to collect missing data or to look more closely at an area. The way we address that is by allowing these different levels of autonomy to be switched in real time. In one autonomy mode, Hovermap provides pilot assistance. In this mode there is still a human on the joysticks, but our collision avoidance is keeping the drone safe, while our SLAM*-based state estimates are keeping the drone stable even without GPS. In the second mode, the user operates the system by placing waypoints on a 3D pointcloud map which is generated by Hovermap in real-time and streamed to a tablet. Hovermap then self-navigates the drone to the waypoint, avoiding obstacles along the way. An operator can interrupt an autonomous mission and revert to pilot assist mode at any time if needed.”

*SLAM - Simultaneous Localization and Mapping

Safer, More Efficient Data Collection

Drone autonomy has helped improve safety in dangerous inspection environments and has made the process faster and more thorough than manual or terrestrial scanning inspections.

“There’s already a step change in efficiency and safety, going from the manual way of collecting data—whether you're climbing a tower or using scaffolding—to using a drone,” said Hrabar. “When you bring in autonomy, there are further efficiency and safety improvements. You don't need to have the expert piloting skills you would typically need and you can get into spaces that would otherwise be pretty much impossible to reach without the autonomy capability.”

Autonomy helps missions happen the same way, time and time again, regardless of who the operator is. “Autonomy can ensure that the data capture is done in a consistent way,” said Hrabar. “And that’s really important for doing visual asset inspection or mapping.”

He continued, “You'll end up having the same image coverage or GSD [ground sample distance] or point density. If you're then running that data through any kind of analytics engine, it really helps because you can train for specific types of imagery or data quality.”

Speaking of data quality, Emesent works with partners and other service providers to make sure data outputs are compatible with other software tools. Some of the data workflows can be automated, but overall, the goal is that the end customer has a streamlined experience from data capture through analysis. 

Drone Collision Avoidance

Emesent’s Hovermap uses LiDAR for collision avoidance to make sure the drones are safe as they’re autonomously moving through spaces. 

“We rotate the LIDAR to give a full spherical field of view,” said Hrabar. “ an obstacle map in real time using the LiDAR data. We're also running a real-time SLAM algorithm on that data to estimate the position, orientation, and velocity of the drone — and that's how we manage to navigate without GPS.”

Hovermap can “see” even the smallest obstacles with it’s Lidar, day or night. According to Hrabar, “When flying at typical inspection speeds, which is one to two meters per second, Hovermap can detect a wire as thin as one millimeter . This ensures the drone is kept safe while flying around telco towers, power lines, under bridges or oil rigs etc. to map and inspect them.” 

Lidar Assisted Visual Inspection

For asset inspection, drone imagery is used for 3D photogrammetric reconstruction and/or visual inspection. “For 3D reconstruction, good coverage and overlap from the imagery is required,” said Hrabar. “For visual inspection, it's more about making sure you've got coverage at the right ground sample distance to see whatever defect you're trying to achieve. A flight which is optimized for visual inspection might not achieve the best results for 3D photogrammetric reconstruction, so there are compromises if you’re trying to achieve both from the same system or sensors.”

With Hovermap, the rotating LiDAR provides the coverage necessary for 3D mapping. “The spherical LiDAR field-of-view means Hovermap is mapping in all directions so the operator doesn’t need to worry about which way the drone is pointed or how it’s flying. He or she can just focus on getting the imagery required for a visual inspection,” Hrabar said. 

In Living Color

LiDAR doesn’t natively have RGB color associated with it, but Emesent recently launched capabilities to capture video from a camera attached to Hovermap and project the RGB data onto the point cloud to generate colorized maps. Hrabar said. “We have a very sophisticated calibration routine, in which we run the camera and the LiDAR together to solve for the intrinsic and extrinsic calibration parameters. This allows us to accurately project the RGB data onto the point cloud”

A key requirement for imagery and mapping to work together for visual asset inspection is tying them together. “Being able to geo reference whatever’s found in the imagery to the 3D map is challenging,” Hrabar said. “Our calibration for colorization also solves this problem, and we’re able to export images that are accurately geo-referenced to the point cloud.”

Organizations can use this technology to collect data in challenging GPS-denied areas, better analyze that data, and keep humans out of dangerous environments. It’s a gold mine of information.