DOI 10.2478/jok-2018-0063 Roman KAZIMIERCZAK1, Wiesław MILEWSKI1, Zdzisław GOSIEWSKI2, Leszek AMBROZIAK2, Cezary KOWNACKI2
1Air Force Institite of Technology (Instytut Techniczny Wojsk Lotniczych)
2Białystok University of Technology (Politechnika Białostocka)
TOWARDS IMPLEMENTATION OF A FORMATION
FLYING FOR EFFICIENT UAV OPERATIONS
Abstract: A flight of a UAV formation is an efficient way to implement surveillance andreconnaissance operations. The usage of a few UAVs as a formation instead of a single vehicle allows creating a distributed network of sensors, which decreases the duration of flight missions and enlarges a total field of view. From a practical point of view, implementations of formation flights require taking into account several separate aspects of flight of UAV such as a quick take-off of several aircraft, aggregating all UAVs in the same space to create swarm and collective flight of the formation towards the area of a surveillance mission. The paper presents the results of researches and experiments carried out towards practical solutions to those aspects. A magnetic launcher is an excellent appliance to put UAV in the air, and its operation could be repeated quickly. Hence, it is ideal to be used in a formation flight. The leader-follower approach based on two-stage switching control is an effective method to aggregate UAVs in the same space while they are flying over large areas. Whereas, the decentralized control of aerial flocking can be used to achieve a coherent flight of UAV formation, which is able to self-organize. Results from simulations and experiments show the effectiveness of each presented aspect and prove their usability in the implementation of formation flights.
Keywords: Unmanned Aerial Vehicles (UAVs), UAVs formation control and flight dynamics, UAVs swarm, aerial flocking, leader-follower control, UAVs launching systems, magnetic levitation launching systems
1. Introduction
Range of practical applications of Unmanned Aerial Vehicles (UAVs) becomes broader from year to year and it includes missions such as surveillance and reconnaissance purposes, aerial surveys for agriculture, traffic monitoring, pollution control, meteorological data collection, pipeline survey, early fire detection, wildlife population tracking, data collecting for precision farming, etc.[12, 34, 40, 50]. The main limitation of those missions is a duration of the flight, which in the case of electrically propelled vehicles does not exceed several minutes of the flight. The effectiveness of UAVs applications can be increased by the usage of formation flights. A formation of UAVs is able to create a distributed network of sensors to increase both qualitatively and quantitatively the effectiveness of measurements in mentioned tasks, simultaneously maintaining the same flight duration. Moreover, as the inversion of the efficiency problem, formation flights could be useful in drag reduction or energy savings approaches expected to lengthen the flight duration and traversed distances [45]. UAVs formations can be also used as nodes of a wireless network, which forward data from one to another to increase connectivity in the network [3].
The theory of swarms or formations has been extensively studied for many years in various research centres around the world [13, 24, 41, 46, 48]. It results in the coherence of research areas such as swarm intelligence, multi-agent systems, robots’ collaboration, formation control, autonomous agents and sensors networks. In those areas, researchers focus mainly on formation flight aerodynamics [2, 14, 30, 31, 35, 42], analysis of modern control laws [23, 39, 44] or synthesis of control laws like optimal [26], adaptive [4], sliding [11], robust [22, 28], vision-based [7, 15] as well as nonlinear [6], where model predictive controller (MPC) with dynamic inversion was used. One of the impressive examples of a formation is the swarm of nano-quadrotors presented by GRASP Laboratory from the University of Pennsylvania [20]. The demonstration flight is very impressive, but the possibility of outdoor usage and the level of the quadrotors autonomy can be questioned. The control of a formation of non-holonomic robots, which is able to operate in outdoor environments, is more complex and complicated, than in the case of omnidirectional vehicles.
Most of the known algorithms designed for the formation control could be classified to three separate subsets of approaches to the multi-UAVs system, namely, bio-inspired behavioural method [19, 29, 43], virtual structure [5, 25, 27, 32, 36] and leader-follower [13, 47, 49]. The subset of approaches treats a group of UAVs as a flock of birds, which behaves according to flocking behaviours formulated by C. Reynolds [33]. The important advantage of aerial flocking is a
decentralized control, which enables self-organization of the flock achieved by mutual interactions between vehicles. On the other hand, the structure of the flock is not stiff, but even it has a random nature. A serious weakness arises from the necessity of information sharing between all or the nearest flock members, to determine flocking behaviours. This complicates connectivity inside the flock because it should provide adequate throughput, and also flexibility related to a varying number of vehicles that could be in the communication range. The approaches in the second subset allow creating a formation in the form of a virtual rigid structure, where relative locations should be constant as much as it is possible [27]. The rigid structure approach makes it possible to assign concrete roles to different relative locations of UAVs in the formation. On the other hand, the virtual structure requires precise position tracking and synchronization and this makes it susceptible to external disturbances. In the group of approaches, only a peer-to-peer relation between a leader and a follower are applied to create formation [13]. In such a case, this is relatively the simplest approach because interactions and communications are limited only to a pair of vehicles.
Other approaches in the field of multi-UAVs systems are usually combinations of some elements of these three fundamental approaches [5, 19]. Their main purpose is to create an algorithm, which in the slightest will be dependent on weaknesses of each of the mentioned approaches, or it will eliminate them completely. Therefore, this paper describes two different methods, which can be combined to increase the effectiveness of formation flights. The leader-follower method based on two-stage switching control is an effective way to aggregate all UAV in the same space while they are dispersed over a large area after take-off. In turn, the aerial flocking provides a decentralized and scalable control providing self-organization of the formation [19]. Hence, the combination of these two approaches will result in a formation of UAVs which be better suited to practical applications, where there is the sequence according to which UAVs must take-off in a relatively short time, next fly collectively toward a target area, disperse in order to complete individual tasks, and finally aggregate together to return to the home position.
The remaining part of the survey is divided into three sections. The first one presents an overall concept of the combination of leader-follower and aerial flocking methods to implement the formation flight sequence. In this section, also a magnetic launcher designed for small UAVs is presented. Next, two sections are dedicated to each of the applied methods in the concept. In these sections, both simulation and experimental results are presented and discussed. The survey ends with the appropriate conclusions.
2. Formation flight
From a practical point of view, performing a formation flight should be considered as a process starting at take-off and ending at landing. Unfortunately, the most of research concentrates only on the control of the flight assuming that all UAVs are already in the air, skipping the part related to take-off of the UAVs formation. The overall process as a sequence of sub-processes, which is essential for a practical usage of the UAVs formation, is presented in fig. 1.
The first element, which is crucial for the effectiveness of formation flights, is a quick take-off of several UAVs. Ideally, each UAV in the formation has the same capacity and level of a power supply, but it is possible with the assumption that all UAVs perform take-off simultaneously. Otherwise, each UAV will have different levels of a power supply, even before the mission begins. It will be reflected in different flight durability for each UAV, and in the case of UAVs propelled with gasoline engines, also in different weights.
Fig. 1. Formation flight as a sequence of sub-processes
Those differences could have the significant impact not only on the effectiveness of formation flight but it could also affect the algorithms of formation control. Therefore, a method which will allow achieving the shortest time of take-offs series as it is possible, is highly expected. The one of possible solutions is applying electromagnetic launchers. An example of such device designed for small UAV was developed in research center located at Bialystok University of Technology [16, 17]. Another example of electromagnetic launcher can be test-bench developed as part of 7EU FP project GABRIEL [38] in Wroclaw University of Technology, (Chair of Cryogenic, Aerospace, and Processing Engineering). This launcher uses 10 meters length magnetic track and high temperature superconductors in Magnetic Levitation (MAGLEV) suspension system of the sledge [8, 9, 21, 37]. Those devices are shown in fig. 2.
a) b)
Fig. 2. a) Magnetic launcher designed for micro UAVs [16, 17]; b) the high temperature superconductors magnetic levitation track for UAVs take-off and landing developed in Wroclaw University of Technology as a part of the GABRIEL project [38]
The construction of the electromagnetic launcher (fig. 2a) consists of ten serially located copper coils with a ferromagnetic core placed inside. The controlled magnetic field of the solenoids affects the core, what results in the magnetic driving force applied to it. The core is connected by means of a diamagnetic pusher to the carriage, to which the launched plane is attached. The whole system is controlled by the open-source Arduino MEGA platform (ATMEGA2560) with implemented algorithm of feedback control of UAV’s position and acceleration [16, 17]. The electromagnetic launcher is capable of launching the unmanned vehicles with a mass up to 25 kg providing the starting speed of at least 20 m/s. Quick recharging process and readiness for another launch are the most important advantages of the electromagnetic launcher, which make it an ideal way to effectively perform the take-off of the entire formation.
In fig. 3, the test bench vertical layout is shown. The model of magnetic suspension system with superconductor consists of magnetic rails and boxes with superconductors. There was performed verification numerical model by the analysis of the Meissner’s effect [8, 9, 21, 37]. Two variants of magnetic rails architecture are investigated. The first variant of magnetic rail was selected to implement launcher. This variant of suspension generated stable magnetic force during horizontal movement of superconductor. The box has got 3 mm underside. The admissible air gap equals 4 mm. The suspension system generated 14 N for air gap 4 mm per one bulk of superconductor. The second variant of magnetic rail was not selected for implementation. This variant generated smaller forces and generated vibration during horizontal movement of superconductor. This variant of rail was suggested as a brake for the end of runway. The magnetic steps can work as a protection of the end of the runway, which was protected with sledge before slip down from runway. Description of mathematical model of test stand and results of simulations are shown in the Delivery D4.4 [38].
(a) (b)
Fig. 3. GABRIEL test bench ground system cross section (a), and layout (b)
This plant uses high temperature superconductors and magnets arrays to produce levitation force. This type of suspension allows stable, passive magnetic levitation even with no linear velocity of the sledge. Such suspension is made up of the path with two parallel tracks (fig. 3b). Frame of the sledge with proper supports mounted underneath is being placed on the top of the tracks (fig. 3b).
The phenomenon of the magnetic levitation is closely related to the Earnshaw’s theorem saying that there is no configuration of permanent magnets, which allows a stable levitation. Therefore, passive suspension using permanent magnets requires additional stabilization by blocking some of the degrees of freedom or by giving levitating object gyroscopic moment. The Earnshaw’s theorem does not apply to diamagnetic materials. Superconducting materials in the superconducting state, as a result Meissner effect indicate perfect diamagnetism. Meissner effect means that from interior of the material, being in superconducting state, all external magnetic fields are pushed off. Superconductivity is a thermodynamic state of matter and occurs at very low temperatures. Most of the non-magnetic elements exhibit superconductivity at temperatures of a few degrees Kelvin. The real breakthrough in the research was the discovery of high-temperature superconductors with critical temperatures comparable to the temperature of liquid nitrogen. The most common high-temperature superconductor is YBCO (YBa2Cu3O7−δ) with a critical temperature of 92K. Laboratories all over the world are working hard to discover the compounds exhibiting superconductivity at higher and higher temperatures. Currently, the highest critical temperature is 135K (HgBa2Ca2Cu3O8). To date, there is a small number of research projects considering the use of high-temperature superconductors magnetic levitation phenomenon in technical issues, mainly in urban transport problems or unmanned aircraft catapults. Superconductivity arouses a growing curiosity of researchers, mainly because of the promising prospects of technical use.
The interactions occurring between the magnetic track and the superconductor are described by the Lorentz force, according to which the lift force created during levitation is proportional to the intensity and direction of the external magnetic field and the magnetic moment of levitating object. However, we should be aware that the Lorentz force is the classical approximation of magnetic interactions that actually require recourse to quantum
mechanics. Model using, only the Lorentz force does not describe the phenomenon of flux pinning effect, occurring in high-temperature superconductors (these are the second type of superconductors). The most accurate theory describing superconductivity is established in 1957, BCS theory (Bardeen, Cooper, Schrieffer) saying that the current carriers in superconductors are Cooper pairs, which are paired electrons with opposite spins, so that they behave as bosons and move in the matter without resistance. The BCS theory, however, does not explain the phenomenon in superconductors of type 2, which still remain a mystery. The phenomenon of flux pinning has very interesting mechanical properties, which can significantly affect the technical solution.
Magnetic suspension is a system consisting of magnetic rails and the levitating cart. Magnetic tracks, over which the catapult cart is levitating, consist of two parallel lines. Each line of the track is lined with three rows of permanent magnets. In test bench, we used rectangular neodymium magnets polarized in up-down direction. There are two variants of arrangement of the magnets on the track. In a first variant, the magnets are arranged in same polarization rows along the track. Due to the repulsion of magnets surfaces contacting the same poles, this configuration is very difficult to arrange, but it generates the optimal distribution of magnetic field lines. In the second variant of the arrangement of magnets, all mating surfaces of magnets are oppositely charged. This configuration is very easy to arrange, but generated magnetic field lines have less desired shape.
The next step in the formation flight is the creation of a tight group of UAVs, when UAVs are flying individually after take-off, usually maintaining wide spacing between them. Wide spacing between UAVs secures them against collisions before they will be ready to achieve a collective flight. Aggregating all UAVs in the same area, especially if they have limited turn radiuses as it is in the case of fixed-wing UAVs, requires a method to minimise distances between them until the moment when the formation control based on position tracking or flocking behaviours will be possible. A two-stage switching control can be useful here because it assumes different control approaches in relation to the size of spacings between UAVs. The idea of two-stage switching control applied to the leader-follower approach is presented in the next section.
3. Two stage switching control
As it was mentioned, after take-off, each UAV flies individually, keeping safe distance from the others. Only when all UAVs are already in the air, the algorithm of formation control can be switched on. Due to the relatively long distances between UAVs, their limited radiuses of turns and inertia in motion
synchronization, the control based on position tracking fails in controlling UAVs to reach their desired positions behind the leader.
Fig. 4. The scheme of formation control based on relation leader-follower: a) the geometry of formation, b) the scheme of formation controller, c) the framework of simulation system [1]
This issue, found by the scientists of the research center at Bialystok University of Technology, is pretty well described in fig. 5, which shows the hardware-in-the-loop simulation results [1]. In turn, the control algorithm applied in the study, the geometry of formation and the framework of simulation system are shown in fig. 4.
Fig. 5. Simulated trajectories of the leader and the follower with applied control scheme from figure 3 [1]
According to [1], the control scheme presented in fig. 4 uses position displacement eyrot in y axis of the formation frame as error signal eψ for PID
regulator (Formation flight heading controller), whose output uψ corrects the
leader’s heading ψLc as a reference heading for the follower. It means that if the
error is zero, the follower's desired heading will equal the leader’s heading. In turn, position displacement exrot in x axis of the formation frame controls the leader’s
airspeed in the same manner. As it is visible in fig. 4, such control scheme is insufficient to track the desired position of the follower in the situation when the leader takes a turn suddenly about 90 degrees or more. A combination of the limited turn radius of delta-wing UAV and rotation of the formation frame due to the leader’s manoeuvre causes a significant error between the leader’s and follower’s headings while eyrot and related PID output is almost zero. It results in a delay of
the follower’s reactions and growth of position displacement. The solution of the problem is a hysteresis loop, which switches the control between two stages according to the actual distance between the leader and the follower. The switching hysteresis loop is presented in fig. 6.
Fig. 6. The switching hysteresis loop [1]
The algorithm II is the control scheme from fig. 4, which could be denoted as precise control stage, and the algorithm I is responsible for initial guidance on the leader when both distance and difference in the headings is significant [1]. In the algorithm I, the position error ex in the East direction and distance R between UAVs
are used to calculate the bearing angle to the leader as the follower’s desired heading. As it is shown in fig. 6, such two-stage switching between two proposed algorithms of precise and initial guidance is an effective method to initialize the UAVs formation flight.
Fig. 7. Simulated trajectories of the leader and the follower with applied two-stage switching control [1]
The moment of switching between the algorithm I and the algorithm II is presented in fig. 8. According to the figure, this happens when the distance R decreases below a value of 60 meters.
A similar two-stage switching control can be successfully applied to aggregate all UAVs just after take-off to create a formation, which will be able to achieve collective flight. The applied algorithm of position tracking based on errors in two axes of local formation frame is the only one of possible approaches to implementation of the second stage of control addressed to collective formation flight. As it was mentioned in the introduction, the position tracking approaches are sensitive to external disturbances, and they assume a priori the absence of collisions between UAVs, if positions in the formation are tracked precisely. Therefore, an alternative to the leader-follower approach could be aerial flocking based on behaviours inside flocks of birds, which were described by C.W. Reynolds [33]. The approach adopting flocking behaviours to a group of fixed-wing UAVs is discussed in the next section.
4. Flock of fixed-wing UAV
The elementary behaviours found by C.W. Reynolds [33], who was observing the flocks of birds, are as follows: cohesion, repulsion, migration and alignment. The first behaviour i.e. cohesion is the most important, because it provides coherence of a flock. Without cohesion, a collective flight of pigeons or other birds would not be possible. The cohesion behaviour concentrates flock's members around a global or a local gravity center. However, the cohesion should be damped by the repulsion behaviour to avoid collision when all members fly side by side. Therefore, a pair of flock’s members should be separated if they are too close to each other, and the repulsion should be considered separately for each pair of members in the closest neighborhood. The migration is a behavior that is used to guide the entire flock, because it defines a target direction for each flock’s member. The last behaviour aligns speeds and directions of the closest neighbors in the flock.
As it was found in [10], many flocks of birds have a leader or small group of flock’s members, which affect the motion of the flock. This means that the leader or the group possessing the information about food source guides the flock by imposing a direction of the migration behaviour. Such strategy, i.e. the combination of behaviours inside the flock, that has a leader, can be successfully adopted to the area of aerial robotics. This was done by the authors of the work [19], who developed a control algorithm applying only two behaviours in the flock of fixed-wing UAVs with the designated leader. The control algorithm has a hierarchical and decentralized structure. Decentralized, because the control is performed individually by each UAV, and hierarchical, because the leader is a parent UAV in the flock. As it turns out, the balance between the attraction to the global center of
mass and repulsions from all other UAVs in the flock allows achieving collective and coherent flight in multi-UAVs missions. The scheme of developed control based on cohesion and repulsion behaviors combined with the leadership is presented in fig. 9.
Fig. 9. The scheme of the control based on cohesion and repulsion behaviors combined with the leadership [19]
According to the scheme from fig. 9, the leader is the only independent UAV in the flock, who listens to commands from the ground control station. Therefore, the supervision of the flock by an operator of GCS is more convenient than controlling parameters of each UAV separately. The global center of mass is calculated as a weighted sum of coordinates of each UAV including the leader, whose coordinates has the highest weight. This ensures that the gravity center of mass will follow only the leader, not other UAVs. In this way, the migration direction will be imposed indirectly by the leader who decides about motion of the global center of mass. The others will apply cohesion behavior to track the global center of mass and repulsion behavior to avoid collisions between them. The principles of cohesion and repulsion behaviors are presented in fig. 10.
Fig. 10. The scheme of cohesion behaviour (a) and repulsion behaviour (b) [19]
b) GCoM UAVn <DR UAVm A B <DC a) GCoM A >DC UAVn
As it is shown in fig. 10a and 10b, the cohesion behaviour makes UAV flying directly towards the global center of mass only if the distance between them is greater than the specified threshold value. In the case of the repulsion behaviour, two UAV are being repulsed exactly from each other if the distance between them is smaller than a specified safe value. To prove the effectiveness of proposed control in maintaining collective flight, numerical simulations were carried out. The results is in fig. 11.
Fig. 11. Results of the simulations presenting flights trajectories of a flock of five UAVs - subfigures (a), (c). The trajectory of the leader is in blue color (UAV1). On the right side, there are plots of 10 relative distances for each pair of UAVs - (b) is related to (a), (d) with (c
Distances between UAVs in each pair decreases gradually and stabilizes at the level related with the balance between cohesion and repulsion. Distances between the leader and other UAVs are greater than distances between them, which means that leader is flying ahead of the rest of UAVs. The main conclusion is that the balance between cohesion and repulsion is sufficient to achieve collective and coherent flight without applying formal migration and alignment. To verify
UAV1 UAV2 UAV3 UAV4 UAV5 0 1000 2000 3000 4000 5000 6000 7000 0 50 100 150 200 250 300 350 400 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 500 1000 1500 2000 2500 UAV1 UAV2 UAV3 UAV4 UAV5 Time [s] Time[s] X [m] X [m] Y [m] Y [m] di st anc e [m ] Al titu de [ m ] Al titu de [ m ] di st anc e [m ] a) b) c) d)
flocking behaviour in the real flights, the implementation of the algorithm was prepared on the basis on a testing framework consisting of two small aircrafts. Fig. 12 presents flight trajectories as results of the experiments, which verifies cohesion and repulsion behavior respectively. In the experiments, the leader was motionless and the second UAV was flying around it presenting cohesion or repulsion behaviour only when the appropriate distance conditions were met. Flight trajectories in fig. 12 confirm that the control algorithm was implemented successfully.
Fig. 12. Flight trajectories of UAV, presenting respectively: (a) cohesion behavior – UAV is attracted to the global center of mass lying in the half way between UAV and the leader, and (b) repulsion behavior – UAV is repulsed from the leader. The leader is located at the coordinates 0,0 [18, 19]
According to experimental results from fig. 12, UAV is able to be attracted (yellow part of trajectory) to the global center of mass lying exactly in the half way between UAV and the leader. But due to the limited turn radius, UAV is circling around the stationary position of the leader. In the case of a repulsion behaviour, UAV is repulsed horizontally from the leader’s position only if the distance between them is smaller than 15 meters (yellow part of trajectory). Parts of trajectory in blue that are located nearby the leader are related to the phases of take-off and landing or manual control mode. The control algorithm was active only between “A” and “B” points.
The proposed control algorithm based on the leadership and behaviours of cohesion and repulsion is an effective way to achieve collective flight of fixed-wing UAVs. It is more flexible than the leader-follower approach, because it is scalable and positions of UAVs are determined by random interactions between UAVs, what makes the structure of the formation more resistant to external disturbances and it also allows to self-organize it independently from a number of UAVs. It is also resistant to a loss of any UAV including the leader, whose role can be reassigned to the other UAV even during flight without disturbing the structure of
the formation. Moreover, as the experiment shows, the flock without the leader will circle around its last known position, which will be stationary. Therefore, the implementation of aerial flocking is a more practical approach towards increasing the effectiveness of multi-UAVs missions. Its combination with the approach of two stage switching control as the method of UAVs aggregation, and quick take-off utilizing electromagnetic launcher is a big step towards more available and effective applications based on formation flights.
5. Conclusions
Practical application of formation flight requires to consider different aspects of the UAV usage, not only the formation of control algorithms, which are mostly verified by synthetic numerical simulations. To significantly increase the effectiveness of applications of formation flights, the problem should be treated as a multi-step process, which starts from UAVs take-off and ends at landing. Quick take-offs of several UAVs achieved by the usage of electromagnetic launchers ensure that vehicles will maintain similar flight duration capability. Two-stage switching control is able to implement two different algorithms dedicated to two consecutive stages respectively, i.e. the stage of UAVs aggregation that sets initial conditions to start a collective flight and the stage when UAVs are achieving the collective flight towards a target area. It should be pointed out that in case of a close (or tight) formation flight, the second stage algorithm of the switching controller should include the aerodynamic effects (additional aerodynamic interactions created by the upwash and sidewash from the lead aircraft and changes in drag, lift, and side force on the follower aircraft) and their impact on the follower dynamics. Thus, in close formation flight, the reasonable approach to control system in the second stage might be a model based control with inclusion of aerodynamic coupling effects as a function of streamwise, spanwise, vertical distance, leader wing parameters like incidence angle, dihedral angle, aspect ratio and taper ratio). In the case of loose formation flight performance (relative distances between leader and follower are greater than the fivefold UAV wingspan), the mutual aerodynamics’ couplings can be omitted. In that situation, the advisable solution is the control of the aerial flocking algorithm. Flocking with a leader is effective in the stabilization of formation flight in a random way, oppositely to a rigid relation in leader-follower approach. The combination of these three described aspects related closely to the formation flight, could be a significant step towards the implementation of formation flying for the efficient operation.
6. References
1. Ambroziak L., Gosiewski Z.: Two stage switching control for autonomous formation flight of unmanned aerial vehicles. Aerospace Science and Technology, 46, 2015: 221-226
2. Bangash Z. A., Sanchez R. P., Ahmed A.: Aerodynamics of Formation Flight, Journal of Aircraft, Vol. 43, No. 4, July–August, 2006
3. Basu P., Redi J., Shurbanov V.: Coordinated flocking of UAVs for improved connectivity of mobile ground nodes. Proceedings of the Military Communications Conference MILCOM 2004, Monterey, CA, USA, 31 October - 3 November 2004, IEEE Press: 1628–1634.
4. Boskowic J.D., Sun Z., Song Y.D.: An adaptive reconfigurable formation flight control design. Proceedings of the American Control Conference, 2003: 284–289
5. Cai J., Sun S. Wu.: UAVs Formation Flight Control Based on Behavior and Virtual Structure. AsiaSim 2012, Communications in Computer and Information Science 2012: 429-438
6. Cheng Z., Necsulescu D.S., Kim B., Sasiadek J.: Nonlinear control for UAV formation flying. Proceedings of the 17th World Congress of the International Federation of Automatic Control. Seoul, Korea, July 6–11, 2008: 791–796
7. Ciesluk J., Ambroziak L.: Vision system for formation flight of unmanned aerial vehicles. Group Flights and Launchers for UAVs. Ed. Z. Gosiewski. Bialystok University of Technology, 2013: 370–386 (in Polish)
8. Falkowski K., Sibilski K.: Magnetic Levitation System for Take-off and Landing Airplane –Project GABRIEL. Proceedings of the 2013 CONSOL Conference in Rotterdam, 2013,
https://www.comsol.com/paper/download/182061/falkowski_paper.pdf
9. Falkowski K., Sibilski K.: System take-off and landing with magnetic suspension – project GABRIEL. RUTMech, 30, 85(3/13), (DOI:10.7862/rm.2013.23), July-September, 2013: 249-258
10. Flack A., Ákos Zs., Nagy M., Vicsek T., Biro D.: Robustness of flight leadership relations in pigeons. Animal Behaviour 86(4), 2014: 723–732
11. Galzi D. Shtessel Y.: UAV formations control using high order sliding modes. AIAA Meeting Papers Online. http://arc.aiaa.org/doi/pdfplus/10.2514/6.2005-6367, (DOI: 10.2514/6.2005-6367), 2005: AIAA 2005-6367 TP
12. Giulietti F., Pollini L., Innocenti M.: Autonomous formation flight. IEEE Control Systems Magazine, 2000, 20(6): 34–44
13. Gosiewski Z., Ambroziak L.: UAV autonomous formation flight experiment with Virtual Leader control structure. Solid State Phenomena. 198, 2013: 254–259 14. Iglesias S., Mason W.H.: Optimum spanloads in formation flight. AIAA Meeting
Papers Online. http://arc.aiaa.org/doi/pdfplus/10.2514/6.2002-258, (DOI: 10.2514/6.2002-258), 2002: AIAA 2002-0258 TP
15. Johnson N., Calise A.J.: Approaches to vision-based formation control. Proceedings of the IEEE Conference on Decision and Control, 2004
16. Kondratiuk M., Gosiewski Z.: Laboratory and field tests of the magnetic coil launcher for micro aerial vehicles. Scientific aspects of unmanned mobile vehicle (Eds. Koruba Z., Krzysztofik I., and Stefański P.), Kielce University of Technology, 2014: 89-107 17. Kondratiuk M.: Concept of the magnetic launcher for unmanned aerial vehicles of mass
up to 25 kg. Proceedings of the 10th International conference: Mechatronics Systems and Materials: MSM’2014. Opole, 2014
18. Kownacki C., Ołdziej D.: Fixed-wing UAVs Flock Control through Cohesion and Repulsion Behaviours Combined with a Leadership, International Journal of Advanced Robotic Systems, 13:36, 2016, (DOI: 10.5772/62249)
19. Kownacki C., Ołdziej D.: Flocking Algorithm for Fixed-Wing Unmanned Aerial Vehicles. Advances in Aerospace Guidance, Navigation and Control. Springer, 2015: 415-431
20. Kushleyev A., Mellinger D., Powers C., Kumar V.: Towards a swarm of agile micro quadrotors. Autonomous Robots, 35(4), 2013: 287-300
21. Ładyżyńska-Kozdraś E., Sibilska-Mroziewicz A.M., Falkowski K.: Investigation of mechanical properties of a rigid body in magnetic levitation state. Challenges of Modern Technology, 6(1), e-ISSN 2353-4419, http://journal.young-scientists.eu, 2015: 23 – 27
22. Li B., Liao X.H., Sun Z., Li Y.D., Song Y.F.: Robust autopilot for close formation flight of multi-UAVs. Proceedings of the Thirty-Eighth South-eastern Symposium SSST’98. IEEE System Theory, 2006: 294–298
23. Li C.: Decentralized cooperative control for multivehicle without velocity measurement. Proceedings of the 48th IEEE Conference on Decision and Control and 28th Chinese Control Conference, 2009
24. Linorman N.H, Liu M.H.T.: Formation UAV flight control using virtual structure and motion synchronization. Proceedings of the American Control Conference, 2008: 1782–1787
25. Low Ch.B., Ng Q.S.: A flexible virtual structure formation keeping control for fixed-wing UAVs. 9th IEEE International Conference on Control and Automation, Santiago, 19-21 December, 2011: 621-626
26. McCammish C., Pachter M.D., Azzo J.J., Reyna V.: Optimal formation flight control. AIAA Meeting Papers Online. http://arc.aiaa.org/doi/pdfplus/10.2514/6.1996-3868, (DOI: 10.2514/6.1996-3868), 1996: AIAA 96-3868 TP
27. Norman H.M., Le Liu, Hugh H.T.: Formation UAV Flight Control using Virtual Structure and Motion. Proceedings of the American Control Conference, June 11-13 2008, Seattle, USA: 1782-1787
28. Pilz U., Popov A.P., Werner H.: Robust controller design for formation flight of quad-rotor helicopter. Proceedings of the 48th IEEE Conference on Decision and Control and 28th Chinese Control Conference, 2009
29. Quintero S.A.P., Collins G.E., Hespanha J.P.: Flocking with Fixed-Wing UAVs for Distributed Sensing: A Stochastic Optimal Control Approach. Proceedings of the American Control Conference (ACC). 17-19 June 2013, Washington DC: 2025-2031 30. Ray R. J., Cobliegh B. R., Vachon M. J., Clinton St. J.: Flight Test Techniques Used
to Evaluate Performance Benefits During Formation Flight, NASA/TP-2002-210730, Aug. 2002
31. Ray R. J., Cobliegh B. R., Vachon M. J., Clinton St, J. (2003): Flight test techniques used to evaluate performance benefits during formation flight, NASA/TP-2002-210730, 2002
32. Ren W., Beard R.W.: Virtual structure based spacecraft formation control with formation feedback. AIAA Guidance, Navigation, and Control Conference and Exhibit, Monterey, California, August 5-8, 2002: AIAA 2002-4963TP
33. Reynolds C.W.: Flocks, herds and schools: a distributed behavioral model. Proceedings of ACM SIGGRAPH ’87, ACM SIGGRAPH Computer Graphics. Anaheim, USA, 27-31 July 1987. ACM Press: New York, USA
34. Saari H., Aallos V-H., Holmlund Ch., Mäkynen J., Delauré D., Nackaerts K., Michiels B.: Novel hyperspectral imager for lightweight UAVs. Proc. SPIE 7668, Airborne Intelligence, Surveillance, Reconnaissance (ISR) Systems and Applications VII, 766805 (April 24, 2010); DOI:10.1117/12.850091
35. Schumacher C.J., Singh S.N.: Nonlinear control of multiple UAVs in close-coupled
formation flight. AIAA Meeting Papers Online.
http://arc.aiaa.org/doi/pdfplus/10.2514/6.2000-4373, (DOI: 10.2514/6.2000-4373). 2000: AIAA 2000-4373 TP
36. Shao Z., Zhu X., Zhou Z., Wang Y.: A Nonlinear Control of 2-D UAVs Formation Keeping via Virtual Structures. Intelligent Robotics and Applications. Lecture Notes in Computer Science 8917, 2014: 420-431
37. Sibilska-Mroziewicz A. M., Ładużyńska-Kozdraś E., Falkowski K., Wolski K., Cedro W., Skalski A.: Experimental measurements of levitation forces generated by high-temperature supeconductors in magnetic fiels. Advances in Intelligent Systems and Computing, 393. Advanced in Mechatronic Systems (Eds. R. Jabłoński, and T. Brzeziński), Springer, ISSN 2194-5357, (DOI: 10.1007/978-3-319-23923-1), 2015: 255 – 260
38. Sibilski K., Falkowski K., Wróblewski W., Majka A.: Ground System Specification and Development to Study the GABRIEL Concept; Deliverable D4.4 Integrated Ground and on-Board system for Support of the Aircraft Safe Take-off and Landing – GABRIEL, EU project number 284884, July 2014
39. Stipanowi´c D.M., Inhalan G., Teo R., Tomlin C.J.: Decentralized overlapping control of formation of unmanned aerial vehicles. Automatica, 2004(40): 1285–1296
40. Tokekar P., Hook V., Mulla D., Isler V.: Sensor planning for a symbiotic UAV and UGV system for precision agriculture. Proceedings of the International Conference on Intelligent Robots and Systems, IROS, 2013: 5321–5326
41. Turpin M., Michael N., Kumar V.: Decentralized formation control with variable shapes for aerial robots. Proceedings of the IEEE International Conference on Robotics and Automation: 2012: 23–30
42. Vachon M. J., Ray R. J., Walsh K. R., Ennix K.: F/A-18 Performance Benefits Measured During Autonomous Formation Flight Project, NASA/TM-2003-210734, Sept. 2003
43. Vásárhelyi G., Virágh Cs, Somorjai G., Tarcai N., Szörényi-Nepusz T.T., Vicsek T.: Outdoor flocking and formation flight with autonomous aerial robots. Proceeding of the IEEE International Conference on Intelligent Robots and Systems (IROS 2014), IEEE/RSJ International Conference., Chicago IL, 14-18 September 2014: 3866–3873 44. Wan S., Campa G., Napolitano M.T., Seanor B., Yu Gu.: Design of formation control laws for research aircraft models. AIAA Meeting Papers Online. http://arc.aiaa.org/doi/pdf/10.2514/6.2003-5730. 2003: AIAA 2003-5730 TP
45. Weimerskirch J.M., Clerquin P.Y., Joraskova A.S.: Energy saving in flight formation. Nature 413(6857), 2001: 697–698
46. Williamson W.R., Abdel-Hafez M.F., Rhee I., Song E.J., Wolfe J.D., Chichka J.D., Speyer J.L.: An instrumentation system applied to formation flight. IEEE Transaction Control Systems Technol. 15(1), 2007: 75–85
47. Xingping Ch., Serrani A., Ozbay H.: Control of leader-follower formations of terrestrial UAVs Proceedings. 42nd IEEE Conference on Decision and Control, 9-12 December 2003: 498-503
48. Yu B.C., Dong X., Shi Z.Y., Zhong Y.S.: Formation control for quadrotor swarm system: algorithms and experiments. Proceedings of the 32nd Chinese Control Conference, 2013: 7099–7104
49. Yun B., Chen B.M.K., Lum K.Y., Lee T.H., A leader-follower formation flight control scheme for UAV helicopters. IEEE International Conference on Automation and Logistics, 1-3 September 2008: 39-44
50. Zhou G., Li C., Penggen C.: Unmanned aerial vehicle (UAV) real-time video registration for forest fire monitoring. Proceedings of the Geoscience and Remote Sensing Symposium, IGARSS’05, 2005:1803–1806
