Dinosour is the drone we developed for the “Teknofest 2025” competition over the course of six months. This blog introduces our UAV.

Design

T-Motor, U11 120KV motors and T-Motor 28.2 folding propellers were used in the design. A storage box was designed to house the batteries. The Pixhawk Orange Cube GPS and LiDAR were mounted on top of this box.

The autopilot system we use is ArduCopter software, which we installed and configured on the Pixhawk Orange Cube.

We use Jetson Orin NX as the companion computer. All image processing tasks are performed onboard, eliminating the need for an external server or ground station. The system operates fully autonomously, except for failsafe scenarios, and does not require any ground connection to carry out its missions. This ensures seamless and independent operation.

Ground Station

As our ground station, we use our custom web based interface, which allows us to adjust vehicle settings and view data from vehicle through a web-based interface.

LiDAR

As a result of our research into obstacle avoidance during missions, we evaluated cameras and LiDAR as the most suitable alternatives. We chose LiDAR because it functions reliably regardless of weather and lighting conditions.

RP LiDAR A2 M8-R4 performs two-dimensional scanning by rotating 360° clockwise within an 8-meter range. It generates up to 4,000 data points per second using the triangular measurement principle.

Image Processing

The YOLO V8 algorithm was used for image processing. The collected visual data was subjected to the labelling process using the ROBOFLOW platform with successful color detection and shape positioning.

Communication

In the vehicle's communication system, the GPS communicates via CAN communication, while LiDAR, Jetson, and RFD use serial communication. The mission mechanism operates through PWM control, and the receiver communicates via SBUS. The communication scheme between these components is shown in the picture.

In the communication tests, we tested the RTK system that we will use on the vehicle. RTK corrects geographical coordinates in real time during flight, allowing the drone to hover accurately even in high winds.

Task Mechanism

A stable and simple task mechanism has been designed to fulfil the desired drop task. The mechanism consists of five bearings, a servo stepper motor, a driver pulley, and PLA material.

When it reaches the desired target, the trigger mechanism activates with a signal from the processor. The dropping rope is designed to be 5 feet shorter than the intended drop height, preventing the vehicle from getting tangled during the delivery.

Test Flights

After completing the labeling process, we analyzed the created dataset using CAB model training on high-performance graphics processors through INS services. This approach significantly accelerated the model's training process. We thoroughly analyzed and evaluated the trained model's performance and accuracy rates. As a result of this evaluation, the necessary steps for improving the model were determined. Color detection was successfully performed on masked images using the CIN algorithm, enabling the effective classification of shapes and letters.

Dinosaur performed test flights as a drone with Advanced flight capabilities ,intelligent obstacle avoidance, Target imaging and target accurate cargo delivery. During test flights, both manual and autonomous flight abilities were demonstrated, with the drone flying autonomously for 5+ miles. Dinosaur is able to climb and stay above 75 feet when more than 200 feet from the runaway.

Aircraft Specifications

• Flight Time : 32 minutes
• Battery Capacity : 32000 mAh
• Battery Type : 12S Li-Po
• Size Including Propellers : 239.78 cm
• Propeller Size : 71.8 cm
• Arm Span : 177.8 cm
• Weight with Payload : 19.2 kg
• Aircraft speed: 54 km/h