GPS and IMU Replacement Using Stereovision in Small Unmanned Aerial Vehicles

           

            GPS is not effective indoors, and measuring systems like LiDAR are too big to be used on some small Unmanned Aerial Systems (sUAS) according to the developers of the Centeye (Barrows, 2016, 0:01-0:28). The idea of the Centeye is to use two light cameras for stereo vision on all four sides of the quadcopter to create a full 360º field of view on which measurements of objects can be made (Centeye, n.d.a). The cameras used on the visual system are less than a 1cm (0.4in) long and weight less than 0.2 grams (0.000441lbs), so they can be installed on Nano quadcopters weighting in at 38 grams (0.0838lbs) (Centeye, n.d.a). While the small format CMOS sensor (Centeye RockCreek™ vision chips) is the heart of the Centeye, the whole sensor is made up of two of the trademark light sensors fitted with 150º field of view lens, a Laser, and two Infrared LED to create stereovision, optical flow, pulsed light, and a Laser ranger (Barrows, 2016, 0:30-1:00), (Centeye, n.d.a). The Centeye sensor can be mounted in four locations to offer full surround vision.

             GPS uses multiple satellite signals to triangulate a location by solving the time it takes to receive the different signals, and Inertial Measurement Units (IMUs) track the accelerations and the movements to follow the movements of a vehicle which can determine a final position by knowing the initial location (Liu, 2017). These methods have their flaws, GPS does not work indoors, or small spaces, and IMU cannot track drift produced by the air motion creating errors. Instead of relying on satellite signals to triangulate location, or tracked movements of the aircraft, Visual Odometry, or V-Slam uses the live changes on the picture obtained by the cameras to calculate where the aircraft has travelled (Liu, 2017). This method compares the current picture with the previous one to locate reference points to track the position of the vehicle. The obstacle is manufacturing a sensor which is small enough to fit in small UAV, and that is good enough to operate drawing low power while obtaining enough information. Doctor Lee from the College of Maryland presented a CMOS sensor on his dissertation capable of enabling this type of measurements to add the ability of visual position to sUAS (Lee, 2016, pp84-100).

            There is not a lot of information available about the technology of the Centeye sensor, or a price, as their model seems to be configure towards fitting custom solutions. The Centeye sensor utilizes a new type of technology which employs IR LED for low light conditions, a LASER to measure distance, and a stereo vision for V-SLAM. The sensor uses a continuous comparison of the current picture with the previous to compare and obtain information that allows it to hover in place, keep track of its location, and identify obstacles. The Centeye solution enables the user to give high level instructions like general direction of travel, while the system uses its cameras to determine a path clear of objects while avoiding hitting blank walls. There is no mention about how it fares with clear windows, and maybe the IR LED can help with this. As this type of technology gets more developed it will make its way into commercial sUAS, and it will expand to other unmanned platforms enabling them to operate indoors, and in narrow spaces as the Centeye sensor is so small it can be fitted to tiny vehicles.

                                         

                                            

Centeye fitted on a Crazyflie platform, total weight 38 grams.                      Complete Centeye sensor

References

Barrows, G. (2016, November 25). Centeye Nano drone with obstacle avoidance November 2016. [Video file]. Retrieved from https://www.youtube.com/watch?v=YTi8bjbZJ4s

Centeye. (n.d.a) Solution for GPS-denied near-Earth autonomy. Retrieved from http://www.centeye.com/small-nano-uas-autonomy/

Centeye. (n.d.b) Vision-based hover in place. Retrieved from http://www.centeye.com/technology/vision-based-hover-in-place/

Lee, T-H. (2016). Enabling hardware technologies for autonomy in tiny robots: Control, integration, actuation. University of Maryland, College Park. Dissertation. Retrieved from https://search-proquest-com.ezproxy.libproxy.db.erau.edu/docview/1814236732/abstract/268A16FED1D044A7PQ/1?accountid=27203

Liu, Y. Gu, Y. Li, J. Zhang, X. (2017, October 13). Robust stereo visual odometry using improved RANSAC-based methods for mobile robot localization. Sensors Journal. 17, 10. Retrieved form https://search-proquest-com.ezproxy.libproxy.db.erau.edu/docview/1965671001?pq-origsite=summon