Solar UAV

Back to Home


I have been participating in the development of solar powered UAV since 2010. The main part of my work is to develop a very light and small autonomous flight control system. Another challenge from controller design perspective is to develop a control law for an aircraft with a very low wing loading since such an aircraft is very susceptible to atmospheric disturbance.

12 hours & 15 minutes of Flight Duration - September 25, 2011

On September 25, 2011, our team performed a flight test with the solar powered UAV. Our aim in year 2011 was 12 hours, and we achieved this goal. The UAV took off at 6:03 AM, flew for 12 hours and 15 minutes, and landed safely at 6:18 PM after the sunset. The distance traveled during this period was approximately 393km. (The aircraft could have flown for about 3 hours using its battery without solar power.)


12 hour-flight in Winter - January 14, 2012

We also flew our solar UAV for 12 hours on January 14, 2012. One of the main reasons that we were able to achieve this long flight in winter time was the MPPT(Maximum Power Point Tracker) that had been developed over the last year by Prof. Hong, Yeh-sun and his student Kwon, Yong-Chul. Since the day time lasts for only about 10 hours in January, we installed four LEDs on the aircraft in order to do the flight test at night time as well.

Flight Control System

The small and light flight control system was developed. The avionics includes a flight control computer(FCC), a wireless modem, and flight sensors such as GPS and AHRS.

A couple of current sensors were also installed in order to see the flow of the electric current in the power system, and an LM35 sensor was installed to monitor the temperature of the electric motor. A separate BEC and an RC relay has been added in order to remotely switch on/off the FCC. The total weight of the developed avionics was under 200 grams including all the associated wires.

Later, we added a device to measure the rotational speed (rpm) of the motor. The following circuit with a phtotoreflector was used for this purpose. This circuit is connect to the USART3 in the above figure.

Flight software was developed using our hardware in-the-loop simulation (HILS) setup. There are a number of computers involved in the HILS including:

Guidance and Control

The nonlinear path-following guidance method is adopted for the aircraft to follow a circular path during the autonomous mode with the inner loop bank angle control. Elevator control input is composed of a trim value plus a perturbation. The perturbation value is for a feedback control to suppress the plane's phugoid mode, and the the trim value was devised such that it can be set manually. Propeller motor power input was used to close the loop for altitude control.

< Preicison Guidance and Control: Flight data while the airplane was flying along a circle making more than sixty laps during a two-hour autonomous flight segment>


One of the benefits of using propeller power (instead of elevator) in closing the altitude loop is to take advantage of the rising air current (=thermal), that can occasionally occurs. When the aircraft comes across such a thermal it gets higher while the power command drops to saturate at the idle (as shown below). Then, the aircraft flies along a helical path since the plane's position is controlled on the cylindrical surface by the nonlinear path following guidance law which is used in the lateral/directional motion.

Wind Estimation and Autonomous Crabbing

The solar powerd UAV is opereated at a very low speed for its maximum duration. Such an aircraft is very susceptible to wind. An efficient way of flying a very low-speed aircraft in a confied area under a strong and steady wind condition, which is often found in the community of model glider pilots, is to have the aircraft crab in zigzags. In this flying skill, the glider is never placed into a tail-wind configuration. Instead, the aircraft is manipulated such that it either flies very slowly in an exact head-wind or it is slightly deviated from the exact head-wind in order to move sideways left or right. By cleverly combining such maneuvers the plane can stay in a confined area.

In order to perform such maneuvers autonomously, a steady wind component in the atmosphere is estimated with GPS velocity only under a novel extended Kalman filter setup. The estimation utilizes the geometrical relation between the air, ground, and wind velocity and the phenomenon that the ground speed increases with a tail-wind and it decreases with a head-wind.

It is also noted in our case that the wind estimation is done without requiring an airspeed sensor. A pitot-static tube with a differential pressure sensor is typically used to measure the airspeed. But if the flight speed is very low, below around 10 m/s, the corresponding dynamic pressure is very low. As a result, most of the differential pressure sensors that are small and light enough do not work in this low pressure range.

Here is a set of fight test data. The figure below shows the flight trajectory. The aircraft took off from around (25m, 5m). Initially, it flew a bit randomly while gaining altitude to about 150m. Then it turned counter-clockwise about three and a half times along a circle, during which the wind estimation was stabilized indicating that a strong steady wind component exists from the south-east, as shown in the next figure. Then autonomous crabbing guidance was engaged, in which the aircraft flew mostly sideways around the center position (-30m, -160m) for more than a hour during the rest of the flight while leaving a number of figure-8 like patterns which are perpendicular to the wind direction, as indicated by the trajectory plot .

< Flight Trajectory >

The figure below shows the estimation variables from the flight test. It indicates that the estimation was stabilized by about t=1000 seconds which corresponds to the point where the aircraft had turned about three and a half times along the circular path. Once after the filter was stabilized the average wind speed estimate was about 6m/s while the airspeed was about 10m/s. The estimated wind direction is displayed in the third plot. In the period after the filter was stabilized the wind direction estimate was 130 degrees in average, which agreed pretty well with the wind data from a nearby Automatic Weather Station of Korea Meteorological Administration.

< Flight Data: Wind & Airspeed Estimation >

The next figure shows the trajectory corresponding to one cycle of the autonomous crabbing from the flight experiment. It also indicates the aircraft attitude along the trajectory by the snapshots displayed every two seconds using the onboard flight data. The wind blowing from south-east corresponds to the direction from bottom to top in the diagram. The crabbing is well indicated in the wide mid-portion in the diagram by the series of snapshots compared to the trajectory line. Also the slow ground speed at head-wind is well indicated at each end by the relatively shorter distances between the aircraft snapshots in this area.

< Flight Data: Autonomous Crabbing >

Autonomous Landing Approach

Autonomous landing approach is also implemented. The primary purpose of this function is to help a safe landing at night hours. From any initial position of the aircraft when this function is engaged, a flight path command is generated that will lead the aircraft to one end of the runway for a landing approach.

A series of flight tests was performed including this one shown below. In this case the aircraft stayed in the first circular path in order to lose an initial altitude. Then it continued to follow the straight line segment and the second circular arc to approach to the runway. Finally the aircraft followed the last straight line segment until the control is taken over to a human pilot for a flare and touchdown.

< Flight Data: Autonomous Landing Approach >


The plane we have been testing in 2011 is a modified version of a commercially available RC airplane, Pulsar from F5 Models. We've done many modifications including the solar cell installation to make it an autonomous solar powered UAV. The solar cell system can deliver up to 70W of power while the electric power required for a level flight is 30W. Our team has a plan to build our own design in the near future.

Last Update - June 13, 2014