Anka is the drone we developed for the “METU VTOL 2023” competition and this blog introduces our UAV.

Design

After extensive research of online resources, aircraft design books, and relevant articles, we determined our aircraft should be a VTOL design focused on payload transport capabilities.

Vertical motors are housed in carbon fiber tubes, with the body designed as a single piece to accommodate these tubes. This one-piece design improves vertical stability by reducing vibration and oscillation. The mold includes 4-degree dihedral and attack angles for optimal production and precision.

Wing design and analysis

NACA 4412 airfoil was selected after XFLR5 program analysis for its superior Cl/Cd ratio at 4° angle of attack, low drag coefficient, and enhanced stability at operational speeds.

For the tail wing, we focused on symmetrical profiles during analysis. Based on the results, we selected the NACA 0012 profile for its stability benefits and ease of manufacturing.

Motors

The VTOL features five engines. Four engines provide the necessary lift during take-off and landing, while the fifth engine generates thrust in horizontal flight mode. During cruising, the four take-off engines are gradually deactivated as the aircraft maintains stable flight using only the single thrust engine. After comparing tilt-rotor mechanisms with the five-motor VTOL design, we chose the more stable five-motor configuration since mission payloads could exert variable forces during transition phases.

Based on research of existing designs and stability considerations, we positioned the fifth motor at the rear. While a front-mounted motor would offer better efficiency, it would create irregular airflow over the fuselage and wings. The rear placement also ensures avionics systems and cameras remain unobstructed during operation.

Propellers

Tests showed each motor provides 5.4 kg thrust in drone mode, using 16-inch propellers. The carbon fiber motor mounts were positioned for optimal balance. To prevent propeller interference with wing function, they were designed to point downward.

Metarial Selection

For the fuselage and wing structure, we chose carbon fiber for its high strength and durability. Carbon fiber offers an ideal combination of low density, affordability, high strength, and flexibility, making it perfect for creating our desired aerodynamic shape. For the wing system connections, we incorporated carbon fiber tubes along with balsa wood.

Mission Mechanism

The mission mechanism uses two fixed parts to carry the payload, preventing any load transfer to the servo. When activated, the servo moves a carrier pipe that releases the payload-holding component, deploying the cargo. The mechanism is integrated flush with the body to minimize aerodynamic drag.

Electronic Components and Subsystems

For vertical take-off: T-Motor U7 V2.0 KV490 motors with T-Motor CF series 16_5.4 propellers. Fifth motor: SunnySky V3 X4125 KV420 with EOLO 15_8 propeller. All motors use FLAME 80A 6-12S V2.0 ESCs and are powered by dual 6S 16000mAh LiPo batteries. Custom power distribution board with fuse protection. PowerHD LF-20MG servos control thrust and flap movement.

Flight control: Pixhawk Cube Orange Plus with Pixhawk Here 3+ Can GPS and RTK for precise location data. Communication through RFD900x RF receiver with 40km range. Image processing via PI CAM V2.1 and RASPBERRY PI 4B. Network connectivity through Bullet M5 AC (VTOL) and ROCKET PRISM AC (ground). Power managed by 5V UBEC. Control via Radiolink AT10II with R12DS receiver.

Ground Station

Our ground station enables safe and easy operation of ArduPilot at flight areas. It features a mini-computer running ArduPilot and an integrated RFD telemetry module for VTOL communication, allowing autonomous flight control and data exchange. The station runs on an external power source for independent operation in the field.

Communication

An image processing system performs the VTOL task autonomously. Images captured during missions are instantly transmitted to the ground station via UDP protocol for processing. The data transmission occurs over a local Wi-Fi network with low latency, enabling rapid communication with the VTOL.

Software

Image Processing

We trained a deep neural network model on our dataset using machine learning methods in Google Colab. The model learned to recognize and classify visual features from the dataset. We then integrated this trained model with the YOLOv8 algorithm to perform object detection and classification.

Flight Simulation

We successfully tested the mission using the Gazebo application, validating the VTOL's vertical landing and take-off capabilities. Gazebo's realistic physics engine and visualization features allowed us to accurately model how the vehicle responds to various Mavlink commands and flight modes.

Flight and Mission Algorithm

After completing 5 full turns for the mission, the VTOL begins moving using data from the location estimation algorithm. It determines and follows the fastest path to the target location. Upon reaching the target, the VTOL switches to Quad mode and aligns itself with the target point identified by image processing algorithms. Once the target point is successfully centered, the VTOL activates its integrated servo motor through PWM signals from the flight control unit (FC). This servo motor then operates a mechanism on the VTOL's underside to perform the payload drop.

Aircraft Specifications

• Flight Time : 30 mins
• Max Speed : 24.5 m/s
• Min Speed : 12.5 m/s
• Cruise Speed : 18.5 m/s
• Weight : 6.5 kg
• Payload : 1 kg
• Height : 0.457 m
• Width : 2.097 m
• Lenght : 1.150 m