Contributions to data transmission at critical moments on unmanned aerial vehicles

José Enrique RODRíGUEZ, Dvid SEGURADO, Mnuel Sá NCHEZ,,José Jvier MARTíNEZ, Rfel GONZá LEZ

a National Institute of Aerospace Technology, Madrid 28850, Spain

b Department of Computer Science, University of Alcalá, Alcalá de Henares, Madrid 28054, Spain

KEYWORDS Critical moments;Data transmission;Data compression techniques;Radio-modem;Recommended Standard 232(RS-232);Unmanned aerial vehicle

Abstract As for unmanned aircraft, the knowledge of the aircraft performance is directly related with the navigation, guidance, and control system programming.Therefore, the measured data in each phase of the flight must be sufficiently precise to obtain a good characterization of aircraft.This article proposes new methods of sending information to ground, which make it possible to know the aircraft behavior accurately, and for this purpose, four contributions have been made for ALO (Avio′n Ligero de Observacio′n, Spanish acronym for Light Observation Aircraft).Currently, the characterization is based on data obtained at ten samples per second, insufficient to acquire detailed knowledge of what happened during the whole flight of an aircraft.As a result of these contributions, many more samples per second of accelerations and angular velocities are obtained at the most critical moments of the flight,such as takeoff or landing.Among the improvements included are data compression techniques, providing references to locate the measured data in time and identifying labels of each parameter.

1.Introduction

INTA (Instituto Nacional de Técnica Aeroespacial) is the Spanish National Institute for Aerospace Technology, and flight tests have been part of INTA’s activity since it was created.With the aim of upgrading such activities and modernizing its facilities, INTA created the Flight Test Area (á rea de Ensayos en Vuelo (AEV) is the Spanish acronym).The AEV is responsible for providing flight test support for all current and future programs, including remotely piloted aircraft system tests,1manned aircraft tests,2–4rocket launches and missile tests.

Remotely Piloted Aircraft Systems (RPAS), also called Unmanned Air Vehicles (UAVs), can be defined as autonomous vehicles5capable of flying by remote control without the necessity of a human pilot.RPAS are very usual in military applications6–7and are of increasing interest in civil applications,8–10such as fire detection and monitoring, searches for people, package delivery and aerial target drones.11–12

INTA has gained a lot of experience over the last years in the design, manufacturing, and application of RPAS for surveillance, reconnaissance and flying targets.Some UAVs have Data Acquisition Systems (DAS) with telemetry and telecommand systems,13others have onboard computers with data transmission via radio-modem, and others have the two systems working simultaneously.14In some specific moments of the flight, more samples per second of certain parameters are needed in order to be able to analyze the real performance of the aircraft15and make decisions in real time,16depending on the recorded results.One of the UAVs developed and assembled by INTA is ALO,a lightweight UAV that provides real-time reconnaissance,17surveillance and target acquisition(see Fig.1 and Table 1).

1.1.Objective

The main objective of this article is to propose several models that improve data transmission to ground on unmanned aircraft at critical moments,taking into consideration limitations such as the absence of a data acquisition system and using transmission by radio-modem.

In order to achieve this goal, the following specific objectives are proposed:

(1) Define the critical moments in the flight of the aircraft.

(2) Propose a model to classify parameters into categories according to the sampling rate required for each of them.

(3) Analyze initial conditions, taking into account bandwidth limitation and available equipment.

(4) Define several models based on different theoretical estimates that allow reduction of the initial limitations.

A further intention is to find the limits of the contributions and study what happens when these limits are exceeded.

Fig.1 ALO.

Table 1 ALO technical specifications.

1.2.Justification

In this type of aircraft, there are no previous studies that take into account a sampling rate bigger than ten samples per second.This is insufficient,especially at the most critical moments such as the takeoff phase18or the landing phase,19where there are parameters with high rates of change.

Therefore, it is necessary to implement new data compression methods that allow the sending of many more samples of those parameters that require it.This implies that more knowledge of the aircraft performance and a better characterization will be obtained.Thus, the navigation, guidance and control system will be programmed correctly.20

1.3.Limitations

These contributions have been developed for ALO and can be extended to another unmanned aircraft called DIANA, conceived as a low-cost target drone21and also developed by INTA.In these small UAVs, due to economic and physical space problems, a DAS cannot be integrated and it has to be equipped with a radio-modem.What is more,it is not possible to store data in real time.

The ALO UAV sends data from separate sensors to the onboard computer, also called MEC (Mo′dulo de Estimacio′n y Control, Spanish acronym for Control and Estimation Module).This information is sent through a RS-23222port to a radio-modem and is finally transmitted to a ground station in the 902–928 MHz band, using unidirectional and omnidirectional antennas.

The Serial Communication Standard RS-232 supports low speeds of transmission and these are lower as the distance increases owing to capacitance effects.In contrast, current data acquisition systems23,24use bit rates of up to 100 Mbit/s,and consequently, requirements of flight test would be perfectly covered and many parameters and samples could be provided.

1.4.Initial considerations

The standard payload is a VNIR (Visible and Near Infrared)imager.25Moreover, ALO has also been used extensively to collect atmospheric measurements, in particular as a platform to take aerosol samples,26to assess their content in DNA chains,and in the framework of studies,to assess the microbial ecology in the atmosphere.

Currently information is sent as follows: 280 data bytes from sensors (in words of 16 bits and 32 bits) by a radiomodem link, and 19 bytes from payload by means of a video link.However, the following article only focuses on the radio-modem link.Fig.2 shows the radio-modem antenna,the antenna for transmitting video signals and the location of the payload.

The critical moments of flight27are specific to each aircraft and each flight and when these take place more information is required in order to correctly analyze the aircraft performance.The critical moments considered for ALO are the takeoff by pneumatic launcher28(see Fig.3) and the landing by parachute29,30and airbags (see Fig.4).

In the same way, sensors installed on the aircraft have the capability to measure a total of 33 parameters, which are the following:

Fig.2 Payload and antennas for data and video transmission installed in ALO UAV.

(1) Accelerations in the x, y and z axes.

(2) Angular velocities in the x, y and z axes.

(3) Angle of attack.

(4) Sideslip angle.

(5) Flight path angle.

(6) Pitch, roll and yaw angles.

(7) Yaw rate.

(8) Magnetic field on three axes.

(9) Ground position and ground speed on three axes.

(10) Latitude and longitude GPS.

(11) Static and differential pressures.

Fig.3 ALO in pneumatic launcher.

Fig.4 ALO in landing phase by parachute.

(12) Vertical speed.

(13) True airspeed.

(14) Barometric altitude.

(15) Height above ground level.

(16) Wind speed on three axes.

These parameters are perfectly defined in floating point according to ANSI/IEEE Standard 75431in 32-bit (4-byte)or 16-bit(2-byte)integer format.Fig.5 shows the main sensors used in the ALO UAV.

The software for digital acquisition via serial port on the onboard computer system is the means of communication used, and it is based on the generation of interruptions onthe microprocessor.The values of each sensor measured at different sampling frequencies are written to a memory buffer and sent to ground with a frequency of ten times per second.

It was decided to group the parameters into three categories.The first one contains those parameters with a very high sampling rate, such as accelerations and angular velocities in the X, Y and Z axis.32Within the second largest group are the yaw rate and the different angles measured (attack, sideslip, flight path, pitch, roll and yaw).The remaining category is made up of the parameters with the lowest sampling rate and few samples per second are needed.Therefore, of the 33 parameters, 20 belong to the first group, 7 the second group,and 6 the third group.This article is based on the last group of parameters.

The number of samples sent to ground is the same for the three types of parameters at the different stages of flight.Hence the ground data reception is linear and uniform.Since much information about parameters with very high sampling rate is required, ten samples per second are not sufficient to accurately know the behavior of these parameters.

Fig.5 Sketch of ALO Instrumentation.

As far as the storage of data is concerned,this is carried out on the ground station, as the type of communication used by the aircraft is based on the management of interruptions,and a possible loss of data could occur in real time.

Fig.6 shows a block diagram representative of the several elements involved in the acquisition and transmission of measured information.

The proposed contributions are based on the following fields:

(1) For the measurement and data acquisition process,IRIG STANDARD 10633and IRIG STANDARD 119 are used.34

(2) For serial communication transmission of data,Recommended Standard 232 is used.

(3) Mechanisms for data compression are based on standards such as the Moving Picture Expert Group MPEG-235and, at a more advanced stage, MPEG-4.36

2.Contributions

The radio-modem works in a frequency band from 902 MHz to 928 MHz.The characteristics of the radio-modem provide a theoretical bit rate of 115,200 bit/s.37However, with the purpose of avoiding interference and allowing a safer remote control, it has been considered that real value is one third of the theoretical value (38400 bit/s).38

The Current Value Table (CVT) is used to store and send data asynchronously39from sensors with different sampling frequencies (see Fig.7).The values measured by the sensors are written to cells at different frequencies,overwriting the current values to the previous values.Knowing the order of the parameters and their size, every 100 milliseconds the CVT receives a clock signal, and the set of all values is sent to ground in one-dimensional array via radio-modem with a transmission format of 8, n, 1 (6 data word, no parity40,41and one stop bit).This process is repeated ten times per second, and when the information reaches ground, the decoding of data is performed.42

2.1.First contribution

In the first contribution, the sending is carried out in a onedimensional array of differential values and using serial data transmission.First, one value in words of 32 bits is sent, and then only the difference from the previous value in words of 16 bits is sent,always taking one bit to identify the sign,a plus sign (+) if the value is higher than the previous one and a minus sign(-)if it is lower.This is done cyclically,hence every ten samples a complete 32 bits sample is sent.Fig.8 shows by means of an example the improvements included in the first contribution compared to the original case.

The same concept of CVT is used.The measured values are written to the corresponding cells at different frequencies depending on the sampling rate of each parameter.However,now a CVT does not exist.There is an array that can store only one value for the lowest sampled parameters, more than one value for moderately sampled parameters and many values for the most sampled parameters.

Nevertheless, there are some disadvantages.There is no label that identifies each parameter.This implies that the sequence of parameters must be scrupulously invariable in both cases: when values from sensors are written to the corresponding cells and then in the positioning within the onedimensional array.Another drawback is that there is no reference to the time in which each parameter is measured,and it is not known if values are equidistant or not.

Fig.6 Block diagram of acquisition and data transmission to ground.

Fig.7 Current Value Table (CVT).

Fig.8 Example comparing original case to first contribution.

Knowing that all parameters have been sampled ten times per second in the original case, in this first contribution, the one-dimensional array contains many more values of the three types of parameters.It was decided that the 20 parameters with the lowest sampling rate(NL)are sampled 20 times per second(SRL).Of these 20 samples, two are sent in words of 32 bits and the rest in words of 16 bits.In addition,the seven parameters with intermediate sampling rate (NI) are sampled 60 times per second (SRI), six of them sent in words of 32 bits and 54 in words of 16 bits.Finally, the six parameters with the highest sampling rate (NH) are sampled 150 times per second (SRH), 15 of them sent in words of 32 bits and 135 samples in differential values of 16 bits.Then, the number of bits per second in each case can be calculated as follows:

As a result, the lowest sampled parameters use 7040 bit/s,the intermediate sampled parameters 7392 bit/s and the highest sampled parameters 15840 bit/s.The sum of all these results is 30272 bit/s, being below the theoretical maximum of 38400 bit/s.

Therefore, the summary of the first contribution is shown as follows:

(1) Low sampling rate: 20 sample/s.

(2) Intermediate sampling rate: 60 sample/s.

(3) High sampling rate: 150 sample/s.

2.2.Second contribution

The second contribution,unlike the previous one,makes use of a two-dimensional array, where the number of rows and columns can be changed, and later information is transmitted in serial.This array includes timestamps to locate in time interval values sent from sensors (see Fig.9), but it is not possible to determine the exact time in which each value was obtained.Both parameters and time references are sent in words of 32 bits and there are no differential values.

As with the previous contribution, the array can store only one value for parameters with small variations,more than one value for parameters with intermediate variations and many values for parameters with large fluctuations in their values.It is also very important that the sampling rate of each parameter does not change during the flight,and in addition,it is necessary that parameters are always placed in the same order because there is no way to identify a change in their order in data post-processing.Therefore,this is an unresolved problem since the previous method did not solve it either.

In this second contribution, all values are sent in words of 32 bits,and 50 time labels are added.This means that every 20 milliseconds an identifying label of the clock value43is sent in words of 32 bits, making a total of 1600 bit/s.Thus, the measured values are between that time (t) and the next time(t + 20 ms).

The parameters with the lowest sampling rate are sampled with the initial frequency of 10 times per second and the 20 parameters are sent in words of 32 bits.Furthermore,the seven intermediate sampling rate parameters are sampled 30 times per second.Finally, the six parameters with the highest sampling rate are sampled 80 times per second.

Consequently, the lowest sampled parameters use 6400 bit/s, the intermediate sampled parameters 6720 bit/s and the highest sampled parameters 15360 bit/s.The sum of all these results (including the 1600 bit/s for time references) is 30080 bit/s, being below the theoretical maximum.

Therefore,the summary of the second contribution is given as follows:

(1) Low sampling rate: 10 sample/s.

(2) Intermediate sampling rate: 30 sample/s.

(3) High sampling rate: 80 sample/s.

2.3.Third contribution

The third contribution contains elements included in the first contribution (concept of differential values) and elements included in the second one (inclusion of time labels in words of 32 bits).In addition, a label that identifies each parameter(T)is implemented,eliminating the disadvantage of the invariable order of the data transmission.

To implement this label,6 bits are reserved for the identification of each parameter.There are 33 parameters,32 of them are identified by labels numbered from 0 to 31 while the remaining parameter will be identified by the absence of a label.In addition, the first value is sent in words of 32 bits and the rest of them are sent in words of 22 bits.This additional improvement means that parameters can have different sampling rates and can even change their frequency at some moments of the flight, subsequently being perfectly recognizable in data post-processing.

As can be observed in Fig.10,a time label is included at the beginning of each row.The sending is carried out in a twodimensional array, where the number of rows and columns can be also changed.Then, data is sent to ground using serial transmission as in previous contributions.

In this third contribution, 100 time labels are added in words of 32 bits.Hence,an identifying label of the clock value is sent every 10 ms,using a total of 3200 bit/s.As in the previous case,the measured values are between that time(t)and the next time (t+ 10 ms).The number of time references per second can be modified.Using more time labels would imply a small decrease in number of samples obtained from the high sampling parameters.

The parameters with the lowest sampling rate are sampled 10 times per second in words of 32 bits.Meanwhile, the seven parameters with intermediate sampling rate are sampled 40 times per second.Of these 40 samples, the first one is sent in words of 32 bits and the other 39 samples are sent in words of 22 bits.The six parameters with the highest sampling rate are sampled 160 times per second.Of these 160 samples, the first one is sent in words of 32 bits and other 159 samples are sent in words of 22 bits.

As a result, the lowest sampled parameters use 6400 bit/s,the intermediate sampled parameters 6230 bit/s and the highest sampled parameters 21180 bit/s.The sum of all the sampled parameters (including the 3200 bit/s for time references) is 37010 bit/s, being below the theoretical maximum.

Fig.9 Method of sending data included in the second contribution.

Fig.10 Method of sending data included in the third contribution.

Therefore, the summary of the third contribution is presented as follows:

(1) Low sampling rate: 10 sample/s.

(2) Intermediate sampling rate: 40 sample/s.

(3) High sampling rate: 160 sample/s.

2.4.Fourth contribution

As in the second and third contributions, the sending is performed in a two-dimensional array utilizing serial data transmission.However,this method provides two variations with respect to the third contribution.On the one hand,it checks if the value that has just arrived is equal to the previous value and,if so,this value is not sent(see Fig.11).This improvement is very interesting in terms of optimizing the radio-modem bandwidth.To make it more practical, it is necessary to implement, on the onboard computer, some margins within which the values are assumed to be equal.In addition,a much more important modification is introduced.This consists of sending the measured values in words of 12 bits, making it possible to increase the number of samples obtained.The first value of each parameter is sent in words of 32 bits and the other values are sent in words of 12 bits,using the concept of differential values.In addition,6 bits are reserved for the identification of each parameter.

Nevertheless, this contribution can also have disadvantages.It is indispensable to study whether these 12 bits can capture the increases or decreases with complete validity because slope overload error can occur.44,45

In the fourth contribution, the first sample of each parameter is sent in a word of 32 bits and the other samples are sent in words of 18 bits, 12 bits corresponding to the measured value and 6 reserved for the identifying label of each parameter.As mentioned above, the current sample is compared to the previous one and written to the corresponding cell only if the values are different.The compared values are considered equal if they make a difference equal to or less than 20%in the case of low sampling rate parameters or equal to or less than 10% in the case of intermediate sampling rate parameters.No difference is allowed in case of high sampling rate parameters.

In the last contribution, 100-time labels are also sent in words of 32 bits, using a total of 3200 bit/s.The parameters with the lowest sampling rate are sampled 10 times per second in words of 32 bits.The first one is sent in words of 32 bits and the other 9 samples are sent in words of 18 bits.Meanwhile,the seven parameters with intermediate sampling rate are sampled 50 times per second.Of these 50 samples, the first one is sent in words of 32 bits and the other 49 samples are sent in words of 18.The six parameters with the highest sampling rate are sampled 241 times per second.Of these 241 samples, the first one is sent in words of 32 bits and the other 240 samples are sent in words of 18 bits.

Fig.11 Example explaining the fourth contribution.

As a result, the lowest sampled parameters use 3232 bit/s,the intermediate sampled parameters 5781 bit/s and the highest sampled parameters 26112 bit/s.The sum of all sampled parameters (including the 3200 bit/s for time references) is 38325 bit/s, being below the theoretical maximum.

Therefore,the summary of the fourth contribution is shown as follows:

(1) Low sampling rate: 10 sample/s.

(2) Intermediate sampling rate: 50 sample/s.

(3) High sampling rate: 241 sample/s.

Fig.12 Accelerations along x-axis in launch phase.

3.Results

Once the four contributions have been exposed, it is intended to obtain the maximum accelerations and angular velocities occurred in both the launch and the landing.The launch phase is carried out through a pneumatic launcher,while a parachute and two airbags are used in the landing phase.The parachute is located at the top of the aircraft and airbags are located at the bottom of the fuselage in order to minimize the difficulty of the landing as much as possible and to avoid any damage resulting from the impact.

The most representative cases are in accelerations and angular velocities along the x-axis.The initial case is taken as the reference point, and then this case is compared to the resulting curves based on the software contributions.The results obtained are shown on the x-axis as time in milliseconds(ms)and on the y-axis as acceleration in G-forces(g)or angular velocity in rad/s.The measured time in both cases is one second.What is more,the legend indicates the number of samples used in each contribution.

3.1.Initial case

Firstly, acceleration in launch phase is shown for the initial case(see Fig.12).The initial curves are made up of 10 samples per second and values have been obtained at the following time: 1, 100, 200, 300, 400, 500, 600, 700, 800, 900 and 1000 ms.Hence, only values at the mentioned points are real and the lines between points are only estimates of what could have happened.

Next,angular velocity in launch phase is shown for the initial case(see Fig.13).The curves are also made up of 10 samples per second and values are obtained at the same time.

3.2.First contribution

In the first contribution, the parameters with the highest sampling rate (angular velocities and accelerations) are sampled 150 times per second,compared to the original case where they were sampled 10 times per second.From a quantitative point of view,this is an important improvement because it is demonstrated that the maximum acceleration in the x-axis of the aircraft in the original case is 7.85g, while the maximum acceleration with the first contribution is 10.605g(see Fig.14).

This improvement allows to know that the real acceleration is greater and make some progress in several aspects.The first of these aspects is to acquire the sensors and equipment necessary to withstand these accelerations.Otherwise, the onboard computer and other additional hardware may be useless once the launch occurs.The second aspect is to define the launcher specifications correctly including all variations detected in angular velocities and accelerations on three axes.However,above all, the greatest improvement is an increase in our knowledge of the dynamics of the aircraft and its behavior.

In addition,angular velocities provide positioning information and,therefore,the capacity of following the programmed course.In Fig.15, angular velocities along the x-axis in the launch phase are shown and,as can be seen,there are also relevant differences from the original curve, which was sampled 10 times per second.

The above-mentioned improvements have been exposed for the launch of the aircraft,but they are also applicable to landing with a parachute, for both accelerations and angular velocities.

As can be observed,in quantitative terms these curves allow to know more precisely the behavior of the aircraft, but in qualitative terms there is an important problem, since there is no time reference.It is not known if the values are equidistant in time or not, hence a positioning error of each value appears on the time axis.

This error is very important in terms of angular velocity because values of angular velocity at each instant are necessary to correct the aircraft actuators and make the plane go as far as it is planned to go.At any instant in time, it is not possible to know exactly or at least with a minimum margin where the aircraft is.

3.3.Second contribution

In the second contribution, 80 samples per second are obtained.In addition, 50 time labels are included, so that a time reference is sent every 20 milliseconds.Therefore, values of angular velocities are located within a 20 millisecond margin and that leads to a significant reduction in the positioning error.Considering that the maximum allowable delay in flight tests is 100 milliseconds, the measured data is good.

Fig.13 Angular velocities along x-axis in launch phase.

Fig.14 Accelerations along x-axis in launch phase of the 1st contribution.

Fig.15 Angular velocities along x-axis in launch phase of the 1st contribution.

As can be observed in Fig.16, the angular velocities obtained in the second contribution are very similar to the original ones.Now, there are time references, the values measured by the sensors are located with greater accuracy and the positioning error on the time axis is considerably reduced.

However, the downside of this contribution is the lower number of parameters sent to ground station compared to the first contribution.Now 80 sample/s are sent instead of 150 sample/s.Both contributions exceed the original 10 samples per second, but the first contribution makes it possible to obtain the values of maximum accelerations more accurately.

3.4.Third contribution

In the third contribution,the software required by the onboard computer is more demanding in terms of number of bits than in the previous contributions.In this case, angular velocities and accelerations are sampled 160 times per second, ten samples per second more than the first contribution, and every 10 milliseconds a time reference is sent, twice as many as in the second contribution (100 time labels per second).As can be seen in Figs.17 and 18,the curves are very similar to those of the first contribution, and no additional information about maximum accelerations and angular velocities is provided.

In contrast to the first contribution,this improvement contains labels for identifying each parameter and data can be sent in a variable order,since values are always identified.Another difference is that parameters with the lowest sampling rate are sampled 10 times per second instead of 20 times per second,while the parameters with an intermediate sampling rate are sampled 40 times per second instead of 60 times per second.

3.5.Fourth contribution

As in the previous contribution, 100 time labels are used and each parameter has an identifying label.In this case, angular velocities and accelerations are sampled 241 times per second,which is the contribution with the largest number of samples.

Fig.16 Angular velocities along x-axis in launch phase of the 2nd contribution.

In the fourth contribution, a slope overload distortion can be noted.This distortion is due to the large dynamic range of the input signal, so when the rate of rise of the input signal is very high, the staircase signal cannot approximate it correctly and the system is not able to provide a good response.To reduce this error, the step height must be increased.However,for very small variations in the input signal,the staircase signal is continuously oscillating around the input signal.This error is known as granular noise.The solution to this problem is to make the step size small.Another way to reduce the slope overload distortion consists of increasing the sampling frequency,but this would increase the required data transmission rate and consequently a larger bandwidth would be needed.Fig.19 shows one example in which the slope overload error is clearly visible.

It is important to add that the intention of the fourth contribution is to study what happens when the limits of the contribution are exceeded.However, this contribution could be perfectly applicable by reducing the requirements of the technique (for instance, not to use words of 12 bits).

4.Conclusions

The contributions made in this article will provide a better characterization of unmanned small aircraft such as ALO,which was originally limited to only 10 samples per second and did not have time references and identifying labels.These new techniques will allow to acquire detailed knowledge of the navigation, guidance, and control system in this type of aircraft without a data acquisition system.In addition, this can also be taken as a reference to provide further information on larger UAVs owned by the Spanish army.

The most important improvements resulting from the four contributions are: the development of new techniques to send the information measured by the sensors in a more optimal way, a more advanced management of the reception buffer in the onboard computer, new labels for identifying each parameter,a better data location in time and the ability to send a larger volume of information.

Fig.19 Accelerations along x-axis in launch phase of the 3rd and 4th contributions.

Fig.17 Accelerations along x-axis in launch phase of the 1st and 3rd contributions.

Fig.18 Angular velocities along x-axis in launch phase of the 1st and 3rd contributions.

In each contribution, different knowledge has been obtained.The first contribution allows to know accurately the maximum accelerations produced in the aircraft at critical moments of the flight, although there are no time references.With the second contribution, it has been possible to locate the measured data in time with higher accuracy.In the third contribution,in addition to the previous improvements, labels for identifying each parameter are added.Finally, the fourth contribution shows that an excessive number of samples per second can lead to a slope overload distortion, which renders the measured data invalid.

5.Future research lines

The study carried out does not imply an end to research in this field.A future research line could be to include new hardware that allows improving the system performance, and thus, to obtain more samples per second of each parameter, increase the frequencies used by radio-modem, and analyze if it is possible to acquire new computers with a lower consumption of clock cycles.

Another research line could be the implementation of new data compression algorithms and their compatibility in unmanned aircraft environments.It would be very interesting to discover better data compression techniques,but above all it would be important to study their influence in real time,where the response time on ground should not exceed 100 milliseconds.It could be very useful to be able to display the measured data in real time, without significant delays or loss.

Another future project is the creation of new development platforms with microcontrollers for ultra-small UAVs.Currently, this type of hardware and software are open source,enabling adaptation to the desired environment in real time,controlling both stabilization and navigation, and avoiding the use of a separate system to maintain stability.These techniques could be included in open-source autopilots of longrange UAVs,since radio-modem does not have a high capacity for data transmission at these distances.For instance, opensource platforms such as Ardupilot,PX4,OcPoC or Paparazzi could be used.46

Finally, the research and study of computational techniques for the detection and correction of errors in data reception could be a good line of inquiry.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

The authors would like to gratefully acknowledge the area of flight test provided by the Spanish National Institute for Aerospace Technology (INTA), especially Miguel Marco,Fernando Lahoz and Jafar Tabrizi, for their help in the instrumentation of the UAV and in the data analysis.