AUVSI SUAS 2022 – ITUNOM UAV TEAM Flight Readiness Review
Coffin has a high wing and conventional tail configuration with large control surfaces so that sufficient maneuverability is achieved for high waypoint accuracy and to easily avoid obstacles. While navigating waypoints having a large link area allows the uav to fly slower and more stable to capture clear and high quality images, pixar, the cube orange is mounted on the body plate with double sided. Vibration, dampening foam tapes alongside a high precision. Gnss system here tree built with cam protocol rtk mode, is used to improve positioning and waypoint accuracy. Optical distance measurement sensor, leader light, helps the uav to land safely. For the communication system rocket, nanostation m5 ac pair is used. These products were choosing considering the conditions such as flight path of aircraft, the angles to be made by the antennas and the radiation patterns rfd900x for telemetry data transfer and rc receiver to enable safety pilot control are integrated into uav sony alpha 6500 controlled by a two Axis gimbal, specifically designed for the odcl and mapping machine, is located right behind the fuselage for the drop mechanism. An electro permanent magnet is preferred over a server mechanism that was used in the previous years. Jetson tx2 is the onboard computer that performs actionably, odcl and subsystem integration over robot operation system for the object, detection, classification, localization machine image, processing, algorithms and deep learning techniques are used with opencv and tesseract ocr libraries algorithms are implemented with python 3 and integrated with ros Kinetic obstacle avoidance algorithm is implemented with mathematical approaches.
Also, controls for flight mechanics are made from the ground station for the new mission mapping web based ptg interface is used to provide 40 minutes of flight time with the safety factor, 26k milliamps hour 10s batteries are used to power the motor, as well as all subsystems To calculate the ugv drop time, projectile motion formulas are taken into account. The antenna tracker system provides orientation on the peach axis and eoxes. The tracker system has its own location and guess the gps location of the aircraft of the mission planner. With background codes, the system finds the required angles to catch the next location of the aircraft with two access in stress 2022. Etunom attempts all of the missions with the coffin and the self confidence levels are indicated in the table. Several types of tests have been conducted throughout the year. All the subsystems are tested. First, with the simulation environments, then, with the test, planes and copters, our team has before integrating all the subsystems into the coffin. We make sure that every one of them works properly. Mechanical properties of the aircraft, such as landing gear, strength, wing strength and trust force have been proved to be sufficient mission. Planner simulation environment is hosted to software development tests and algorithms are optimized before being tested on the coffin. The team conducted 30 autonomous flights so far corresponding to approximately three hours of flight time for stable and accurate autonomous flights. Roll and peach tuning parameters are set by autotune system.
Autopilot learns how the aircraft responds to sharp attitude changes and provides a confidential flight after auto tuning the roll and pitch parameters. The uavs characteristics of waypoint accuracy were tested on different flight paths, and these parameters are again adjusted manually previous years. Waypoint missions were demonstrated many times and accuracy has been improved significantly throughout tests to further improve the waypoint performance of the coffin l1 controller, known as the navigation controller, which produces much more accurate flight pads is tuned manually after the improvements, the attempted 100 waypoints are all Successful hit with an average of 15 feet accuracy, the obstacle avoidance algorithm, is tested. With the mission example given on the github account at first, then, some extra mission maps are created on our own for further testing in the mission tests 20 out of 25 stationary obstacles are avoided, which means the algorithm has an 80 success. Rate missions are demonstrated on both mission, planner simulation and actual flight sony alpha 6500, with 16 to 50 millimeter lens used in the imaging system in a captured image. The aspect length of an object must be at least 20 pixels to be detectable. For this reason, objects of sizes, 11 to 40 inches were used in the tests, and images were captured from 140 170 and 200 feet. Tests showed that out of 24 objects captured from these heights, 24 22 and 18 of them were detected successfully respectively. To prevent changes that may occur due to conditions the shutting mode is set to 1 over 2500 also focus iso and other adjustments are changed during shooting to prevent perspective shifts in the image, a gimbal that will provide stability in the pitch and roll axis has been Designed the detection algorithm is tested on more than 250 images and 80 region of interest for each was found, which means there are many redundant to illuminate them.
The algorithm is improved and the region of interest for image ratio to gives to 13 and 75 percent of total region of interests are found to be correct. Shape, detection algorithm detected 32 of 37 different object, shapes correctly to lower the runtime of algorithm. Multiprocessing was used and reduced by 10 times, tesseract ocr algorithm detected 28 of 37 alphanumerics correctly. An orientation of 26 of these were determined correctly. A color detection model is trained with 800 rgb values, training which, up to 82 percent success with high proof perimeter tuning overall autonomous success of odcl algorithms is found to be 80 percent. The cameras hatchier feature is used to get the exact time. The photograph is taken with this information. The exact location of the uav is extracted when the image was taken. Real location of the object is calculated. Based on that information, the algorithm is tested on 100 different region of interests. Results show that we can localize objects with an average error radius of 13 feet. The mapping path was drawn based on 1200 feet height from the map center and the total of 30 40 50 images were captured from 260 200 170 feet respectively. Also 50 overlap between images is achieved to get similarity. Ptgui interface is used to create maps from images for each altitude. It is seen that the more images the better quality, however, it causes perspective problems. The optimum solution is determined to take 40 photos from 200 feet.
For the ugv system, a 3d printed shock, absorbing structure and the convenient parachute system are designed and manufactured to be able to withstand the drop impact the ugv survived. Thirdly, out after the eight drop tests in the manner of strength, driving tests are conducted to observe the ugvs ability for autonomous navigation. After the improvement of driving system, ugv is finally able to drive the target location within five feet. Distance drop tests are conducted to observe landing accuracy, the dropping algorithm is optimized and the accuracy has fallen between 13 feet after 20 trials, since it numb is a familiar theme with the swast competition. Several competition exercises are held together with the experiences from the past years. Full mission tests were performed with all the possible scenarios and the outcomes were truly satisfactory with all the final subsystems integrated into the uav five full mission tests were performed during these test flights. Problems arising from the interrogation of all systems were detected, for example, airdrop mechanism. Could not be triggered due to the magnetic field generated by all the subsystems working at the same time, proper cabling is done and the magnetic shield is applied to overcome this issue during obstacle. Avoidance gimbal performance was negatively affected. The response time of the gimbal is lowered and the problem is sold thanks to the general and special checklists. We made our preparation process the fastest and minimize the risk of human error. With the importance we attached to strong communication and professionalism, we became a team that makes a difference in the field.
Full mission tests showed us that ethnom is sufficiently capable of performing timeline, autonomous flight waypoint capture, obstacle avoidance, odcl mapping, airdrop and operational excellence missions. Five full mission test results and the estimated score of suez 2022 are in the table.