Throughout this blog, we've aimed to educate our customers and anyone interested in SLAM and autonomous robotics about the sensors that make up a mobile SLAM platform. Prior to the release of Nexys we even went into more detail, describing how each sensor that feeds information to ExynAI is like a human sense or organ, while ExynAI is the brain that intelligently processes the information about its current mission.
One of the most important sensors used to create accurate 3D models is a LiDAR sensor. You can think of it like the robot's eyes. Light Detection and Ranging (LiDAR) technology is a technology whose principle is based on emitting multiple laser pulses and measuring the time they return to a LiDAR receiver. On Nexys, the sensor is then rotated 360º to capture a complete map of the environment. The resulting data is processed to create accurate maps, terrain models, or three-dimensional images of objects often referred to as a point cloud.
Another sensor of interest, and one that we get the most questions from customers and channel partners about is our internal measurement unit, or IMU. IMUs like the one inside Nexys are exceptionally common throughout our digital landscape. There's a great chance that the device you're reading this blog post on contains an IMU! These sensors help devices/robots/SLAM understand how they're moving through 3D space. This could be for something as simple as changing the screen orientation of your phone, or vastly more complex like feeding pose estimates into a Simultaneous Localization And Mapping (SLAM) algorithm for autonomous navigation.
But if you've never used a SLAM platform before, you've likely mounted a LiDAR sensor to a fixed wing plane, drone, or mounted to a vehicle and connected it to an internal navigation system (INS) that will enable the data to be post-processed into a 3D point cloud. An INS inside a fixed-wing drone can provide complete information about the navigation parameters of the drone's movement – heading, pitch, roll angles, acceleration, speed of movement, and coordinates of the drone's location. For an INS system, these precise IMU measurements need to be tightly coupled with GPS coordinates or similar dataset to create a globally accurate 3D point cloud.
In a SLAM system like Nexys, in contrast to a true INS, we use the IMU and inertial information primarily for short term trajectory information, on the order of seconds. This is used to propagate motion over this short period and un-distort scans, etc. But across a large scan, we use LiDAR information, not the IMU, as the primary source of information to maintain global accuracy. We build a map and maintain localization within it. The IMU mostly just helps us put small pieces of the map together rather than the entire thing.
So why not just use the most expensive IMU we can find? Much of the cost of these expensive IMUs and INS systems is in reducing the drift rate of the IMU biases to maintain accuracy over long trajectories. In a SLAM system, however, we are able to continuously estimate these biases as they change through sensor fusion, making driving these drift rates to zero on the physical IMU much less critical.
In addition, we rigidly mount our IMU and LiDAR sensor inside Nexys and tightly calibrate both to further reduce any bias or drift. This ensures Nexys is capable of capturing accurate, feature-rich 3D models while still maintaining its modularity to move from handheld to vehicle- or drone-mounted configurations. Adding extensive calibration steps to integrate an INS system would nullify the cost and speed advantages of using a Nexys for handheld or autonomous mapping while providing marginal to no improvements to map quality.
From the formation of Exyn, we wanted to ensure that our mapping and autonomy algorithms would be able to integrate onto nearly any robotic platform. We didn't want to pigeon-hole our autonomy onto just aerial or just ground-based platforms. Or just specific robotic platforms even, like DJI or Boston Dynamics.
We looked at endless sensor configurations and planned for what sensors would work best for our needs for autonomous SLAM flight and after-mission 3D map creation. There were countless tests, prototypes, almost-there's, and of course a few failures along the way. But throughout our rigorous testing both in our offices and through hundreds of thousands of flights in the field, we've arrived at the sensor package that best suits our product goals and the needs of our customers.
or to a drone for autonomous exploration
Regardless if you're using an INS-based LiDAR or a mobile SLAM platform, large-scale 3D mapping is hard. But survey teams equipped with a Nexys for mobile SLAM data capture can be quickly deployed for a lower price and with better resolution than an expensive INS-based LiDAR solution.
Contact us today to customize a personal demo to experience how Nexys can fit seamlessly into your unique 3D mapping use case.