Unmanned aerial vehicles (UAV), colloquially known as drones, are reinventing businesses and even creating brand new opportunities. Check out our article on Dzone to learn why drones are useful for business and how we develop them for an AgroTech company.

Drones do more than just shoot beautiful videos during travels and other memorable events. They deliver parcels, monitor objects, and can even replace people on deadly missions. Read this article to find out what businesses can use drones for and why, as well as how we help iFarm, an AgroTech startup, address business challenges using drones.

Implementation
The drone consists of a carbon frame, a flight microcontroller with software, brushless motors, engine speed controllers and software, propellers, a battery, peripheral power supply system, lidar and wires connecting all of the subsystems. The flight microcontroller has a built-in accelerometer, barometer, temperature sensors and other sensors that monitor the current consumption and voltage of the battery.

A brief algorithm of the flight microcontroller is as follows:
1. Computer vision algorithms send calculations to sensors and communication interfaces. The flight microcontroller then collects data from sensors and communication interfaces. From this, a complete source data package for the PID controllers of the flight microcontroller firmware is formed. PID controllers generate a control signal to obtain the necessary accuracy and quality of the transient.
2. Based on the data obtained, the 32-bit core of the flight microcontroller calculates the values of the control signals in real time and sends them to the engine speed controllers.
3. Upon receiving the control signals, the engine speed controllers, with the help of firmware, calculate and direct the necessary current from the battery to the motor windings, with the desired frequency. In response to the microcontroller, they send data on the current real speed of the motor and its consumption.
This is only part of the entire drone system. There is also a microcomputer installed on the drone. We are testing various microcomputers of an appropriate size. We have already mastered Raspberry PI, Rock Pi and Nvidia Jetson Nano. The next one is the most productive and technologically advanced in its size, the Nvidia Jetson Xavier NX, which opens up new horizons for the development of intelligence of the drone.

On such microcomputers, we perform calculations for various algorithms of computer vision, odometry and neural networks. Concurrently, we use video streams from cameras along with the sensor data provided by the flight microcontroller. Thus, the microcomputer and the microcontroller help each other stabilize the drone in space. Additionally, with the help of microcomputers, we encode and convert various formats of video streams and solve applied problems, including drone control.

Stage One Results
At the first stage, we wanted to check whether the results were accurate enough when using the concept of marker positioning for "narrow" monitoring tasks. The drone was expected to automatically fall into the aisles between 80 cm wide rows and carry additional equipment on board. We needed to select and test the most efficient component model of the drone, its aerodynamic shape, and check a number of electronic components from Chinese, Turkish and American suppliers. After six months of work, we completed all the tasks. We are pleased with the results and, during the research process, generated many ideas for further improvement upon the design.

Further Plans
As for the monitoring system, we plan to work on the following areas:
1. Continuing work towards markerless positioning. The marker system works well, but we do not see the future of autonomous drones behind it. Drones should be smart enough not to require any external clues that might warp over time. We see a future where drones use the full range of modern technologies exclusively on their own board, such as neural networks, stereo vision, visual odometry, machine learning, and lidars.
2. Developing a workplace for farm employees that allow for the creation of routes and flight plans for monitoring objects, as well as monitoring the results of work done by drones. This will help support the functioning of the marker system.
Develop a system for receiving light loads so that it is mounted in the base of the receiver's standard windows. Thus, we plan to implement a system for receiving light loads and small parcels weighing up to 1 kg without the need to visit the points of receipt of cargo. We do not foresee, in the near future, the widespread use of drones capable of delivering loads heavier than 2 kg. Heavier loads carry a risk to human safety. They require the development of regulatory standards and security systems, which, in the best case, will take 2-3 years. However, getting a basket of strawberries without GMO, which were growing in the garden 5 minutes ago, directly in the window, is a realistic task right now.