The advent of powerful embedded systems, sensors, and communications networks has drawn widespread interest in the use of unmannedaerialvehicles (UAVs) to execute missions with limited involvement of human operators. In recent years, UAVs have been playing an increasingly important role in military reconnaissance and strike operations, border patrol missions, aerobiological sampling, forest fire detection, police surveillance, and recovery operations, to name but a few. In simple applications, a single vehicle can be managed by a crew using a ground station provided by the vehicle manufacturer. The execution of more challenging missions, however, requires the use of multiple UAVs working in cooperation to achieve a common objective. In such missions, a team of vehicles connected by means of a communications network must meet stringent spatial and temporal constraints, while possibly maneuvering in close proximity to each other. In general, success of these multi-vehicle cooperative missions depends on the ability of the fleet to exchange information in a timely and reliable manner and, therefore, the quality of service (QoS) of the supporting network becomes a factor of major importance. In addition, as pointed out in [ 49 ] and [ 55 ], in many scenarios the flow of information among vehicles may be severely restricted, either for security reasons or because of tight bandwidth limitations. As a consequence, no vehicle may be able to communicate with the entire fleet and, moreover, the amount of information that can be exchanged may be limited. Under these circumstances, a key enabling element for the effective execution of multi-UAV missions is thus the availability of cooperative motion-control strategies that can yield robust performance in the face of external disturbances and communications limitations, while ensuring at the same time collision-free maneuvers.
tion strategies are discussed in the context of ground vehicles but are applicable to aerial applications when constraints are not a factor. Ahmadzedah et al applied receding horizon optimization techniques to minimize the time required to persistently cover a designated area . Strictly satisfying image refresh requirements over a long period was found to be uncertain. Gorecki et al  apply model predictive control optimal costs to balance explo- ration, safety, and mission termination specifications. Persistence is not directly considered but could be accommodated with an iterative execution of their algorithm. Performance guarantees for coverage would also require additional development. Nigam and Kroo  develop persistent surveillance policies for single and multiple agent applications. Their results show the merits of basing plans on feasible path lengths rather than Euclidean dis- tances, especially when turn rate capability is relatively low. Their approach to multiple UAS persistent surveillance, which is based on an optimum policy for a single-UAS case, validates the idea of basing team performance on the analysis of a single agent. Acevedo et al present an area partitioning strategy to solve the problem for irregular areas and het- erogeneous UAS . Caraballo et al  generalize the concept using a strategy defined as the block-sharing technique to accelerate convergence to an optimal partition. Vehicle capability in their model is specified by speed and sensor/camera field-of-view; fixed-wing maneuver constraints are not included. Mixed Integer Linear Programming solutions have been developed by How et al  for maximizing coverage but not for persistent applica- tions. Wallar et al  directly address persistence in the context of a reactive planner tailored for agile quad-rotor platforms, so further work would be required to accommodate fixed-wing levels of maneuverability. Finally, in this survey, Cowlagi presents the case for a hierarchical approach to optimal planning between points .
An LPV plant model was first introduced by Shamma and Athans  whereby its dynamic characteristics vary, following some time-varying parameters whose values are unknown a priori but can be measured in real-time and lie in some set bounded by known minimum and maximum possible values. An algebraic manipulation method, e.g. Jacobian linearization [39, 66, 80], state transforma- tion [13, 92], or function substitution , etc., is normally used to derive an LPV model from the original nonlinear model. Moreover, in the literature, there are sev- eral different varieties of LPV models, e.g. the grid LPV model [39, 66, 100, 101], the affine LPV model [8, 7, 10] (or polytopic LPV model), the tensor-product (TP) convex polytopic model [15, 16, 18], etc., these have been introduced for the analysis and gain-scheduled control synthesis which is usually based on sin- gle quadratic Lyapunov function [10, 23] or parameter-dependent Lyapunov func- tions (e.g. parameter-dependent [8, 39, 100, 101], affine parameter-dependent , piecewise-affine parameter-dependent [63, 64], blending parameter-dependent , multiple parameter-dependent Lyapunov functions [65, 66], etc.).
A new method to obtain a flat phase margin with FOPID controller is proposed . This method is applied for the roll control of a UAV using the specifications: gain margin, phase margin and robustness to gain variations of the system with two different approaches. The first approach consists in a limited number of frequency samples of G(s) around of its crossover frequency for the controller design. This approach allows greater computational efficiency than the classical method, which uses a first order plus time delay (FOPTD) model. In the second approach, an approximated open loop system (G 0 (s)) with the same amplitude of G(s) and different phase curve is used. Moreover, two additional controllers are used: an IOPID controller and a classical FOPID. For the performance evaluation of the controllers, uncertainties in the aerodynamic parameters of the UAV are added. The simulation results in MATLAB demonstrate that the new FOPID controller is more robust than the other controllers in closed loop.
This chapter will address multi-agent coverage problems involving agents with second order dynamics. We will consider a partitioned region for which the agents must provide coverage. These problems can be related to many real-life applications such as search and rescue, deployment of mobile network access points, and surveillance. To develop a control law for the agents we will require a property of Voronoi cells, that is derived in this chapter, which is not presently available in the literature to the best of the author’s knowledge. This chapter is structured as follows. The preliminaries on Voronoi diagrams and Reynolds Transport Theorem are reviewed in Section 4.1. The dynamics of the area of a Voronoi cell is derived geometrically, and using Reynolds Transport Theorem, in Section 4.2. Section 4.3 derives the dynamics of a Voronoi cell’s center of mass using geometry and Reynolds Transport Theorem. Finally, a coverage problem is addressed using the results on Voronoi cells in Section 4.4.
The second case builds on the work of the first case, which solved a point-to-point collision- free path, and expands the environment to an obstacle-free environment with numerous target locations. The trajectory is segmented from waypoint to waypoint while still maintaining its continuity and satisfying the kinematic constraints of the UAV. For a multi-waypoint path planning problem, it requires the UAV or robot to perform the specified speed, position, and sequence computed by techniques, such as heuristic algorithms or polynomial methods, in order to reach all waypoints [Davies and Jnifene (2006); Hashim and Lu (2009); McGee and Hedrick (2006)]. Another extension of this work is evaluating decisions of the unknown sequence to pass through all waypoints from a given beginning location. However, it is difficult to decide the se- quence unless it is pre-defined or evaluated separately from the dynamic system. New concepts for decision making have been proposed, for instance, hybrid optimal control system designed with switching phases while the vehicle is conducting different situations [Ross and D’Souza (2005); Soler et al. (2010)]. To handle the decision making of visiting sequences, along with the kinematics of the UAV, a set of hybrid decision variables, including both binary and continuous variables, is introduced. The problem is formulated into a form of general/nonconvex QCQP problems, and the flight path and the sequence can be solved simultaneously with an iterative rank minimization (IRM) algorithm.
It is usual practice trying to summarize the UAV state with a single variable that takes values from a finite set. For example, the set of possible states could be defined as f landed , flying g. As we wanted this concept of state to be also abstracted from the autopilot specifics, we decided to give to UAL the responsibility of both defining the set of possible states and keeping updated the current state of the UAV. In the first software ver- sions, we tried to implement the UAL state update with a simple state machine. Initial state was landed and each of the following function calls caused a logical change in state. For example, calling the takeoff func- tion changed state to flying , and then calling the land function turned the state again to landed . However, this approach did not work properly in field tests. Any- time the system needed to restart while flying, or when the human safety pilot had to take control and take off or land, the UAL state did not correspond to reality. This led us to the conclusion that it was more realistic to estimate the state at each update instead of keeping a state machine running. As the function in charge of estimating the state has to deal with some autopilot specifics, it is implemented at back-end level.
the dynamics . By having multiple networks and a dynamic switching technique, accurate modelling for different flight conditions can be achieved. The concept of multiple neural network models for adaptive control was introduced by Narendra in . In the past, multi- networks have been used for linear systems or offline modelling that are mainly based on numerical simulations –. In this work a real-time implementation of multi-networks for a UAV is presented. A well trained offline model based on previously collected flight data is used in conjunction with a simpler online model to achieve better accuracy. The switching between the online and the offline models is carried out based upon a suitable criterion. Two different selection criteria are considered for the dynamic selection process. In the first case, the instantaneous errors between the predicted outputs and the actual outputs are considered for selection. In the second case, the past performance of the two networks are also taken into consideration. A weighted sum of the instantaneous error and the mean square error (MSE) is used for selection. Each model uses a multi-input multi-output architecture. Modelling is performed with MLP networks based on the autoregressive technique with exogenous inputs (ARX) forming recurrent neural networks (RNN) introduced by Ljung in . A novel training method is adapted for the online model where the network is trained with small batches of data and the weights from the previous batch are retained in memory . The retraining of the online network is carried out only when the prediction error increases beyond a certain threshold. The online and offline identification techniques have been flight tested repeatedly. The multi-net model has been validated with a real-time hardware in the loop (HIL) simulation technique. The viability of the proposed technique in real-time is proved using the HIL simulation, hence allowing it to be implemented on flight tests.
Another highly influential survey paper, written by Patton in 1997 , classifies the FTC prob- lem as a complex control system requiring inter-control-disciplinary information and expertise. Patton argues for new controllers that can tolerate component malfunctions whilst maintaining desirable and robust performance and stability properties. Patton  concurs with Eterno et al.  that the main requirement of an FTC system is either the maintenance of an acceptable level of performance, or graceful degradation following a malfunction. For a real time application Patton suggests the comparison of several methods on the basis of cost, robust stability, degree of predictability of the behaviour of the system and whether or not the system could degrade gracefully without loss of life/injury and/or significant economic loss. Other important factors in the decision making process include computational burden as well as the complexity of the system as a highly complex system could decrease overall system reliability. Patton comes to the conclusion that the main objective of fault tolerance should be the design of a controller able to guarantee stability and satisfactory performance, not only during healthy operations but also during component malfunctions. Such structures are referred to as control loops that possess loop integrity or reliable control. Hence FTC is a strategy for reliable and highly efficient control law design. According to Patton, research in FTC during the 1970s and 1980s has concentrated on:
It can be seen that the system’s output phase lag, which is primarily attributed to latency in image capture and feature extraction, increases with increased input frequency as shown in the accompanying Bode plot shown in Figure 7.3. Figure 7.4 shows similar data for the gantry in isolation, while Figure 7.5 is the corresponding Bode plot. It can be seen that the the arm alone has about half the bandwidth of the gantry alone. Additionally, rise time of the arm is approximately twice that of the gantry. This is an important consideration for the application of partitioning where it is desirable for the faster DOFs to be used as sensors for the system’s slower DOFs. This was not the case for this evaluation, but the partitioning algorithm was evaluated nonetheless, and the combined system performed better than the individual components, having a higher bandwidth and lower rise time as shown in Table 7.1 and Figures 7.6 and 7.7. The success of the partitioned control of this combination of DOFs is attributed to the lack of positioning perturbations to the gantry. Had the position of the gantry been randomly perturbed, it is expected that the manipulator, having a lower rise time than the gantry, would not have been able to adequately compensate.
Unmannedaerialvehicles have seen a rapid growth in use for recon- naissance applications with a wide range of vehicle types and capa- bilities fielded. Most current UAV types require a fixed base-station to up-link way-points to the UAV, which have been determined by a human operator, or possibly by path-planning software hosted by the base-station. As the use of UAVs becomes more commonplace, there is a requirement to de-skill the operation of UAVs to allow untrained operators to use these systems in the field with a minimum of ground equipment. This will require much of the path-planning capability to reside on the UAV with the operator up-linking only high-level goals, which must then be autonomously executed. Such autonomous path-planning must operate in near real-time, must be computationally efficient and must be validated to ensure that the UAV safely achieves its goal.
Another finding was that, though the simulation proved otherwise, angular acceleration can be determined from flight data using numerical methods. The actual gyroscope measurements show less noise than that simulated in the non-linear model and thus could be differentiated. In summary, this thesis showed the processes involved in obtaining aircraft parameter estimates from flight data. The different problems associated with obtaining these estimates were discussed and recommendations for alleviating or minimizing these problems were presented. This thesis thus provides a basis for further research into the implementation of aircraft parameter estimation for real-time applications within a Fault Tolerant Framework.
This section introduces the dynamic real-time rerouting problem applied to UAVs, provides a generic formalization, and presents the solution created for this thesis. The word route, as opposed to path, is used because it is the term most widely used by the UAV community, Dynamic inflight UAV rerouting is a special case of online mapf applied to UAVs. In the UAV ecosystem, unexpected events may occur in the airspace, forcing autonomous UAVs to deviate from their designated routes. A UAV may experience engine failure and need to safely maneuver to a nearby landing location. A medical helicopter may need to occupy the same route as a UAV. Inclement weather or police activity may cause a large portion of the airspace to be temporarily closed. These are all events that dynamically add an obstacle to the mapf problem at runtime. The terms obstacle and no-fly zone will be used interchangeably. A UAV experiencing engine failure can be seen as an obstacle because it will not be forced to change its path and therefore must be avoided by other UAVs. Every UAV has an operation. An operation consists of a start location, s i , service location,
One area where UAVs are most useful is in search and rescue missions. Search and rescue is traditionally done in the wilderness to find people that have gone missing, such as hikers. Normally this is a very costly and dangerous procedure, as rescuers need to trained and compensated. It is dangerous because hikers are normally lost in hazardous and hard to get to terrain. Sending in people after them is risky as it exposes them to the same risks that the hikers have been subjected to. UAVs solve this issue by being sent to fly over the terrain that is being searched. On-board cameras can get visual images of the area below and run image processing algorithms on the video images to detect any background disturbances. The algorithm can detect whether the disturbance detected is a human and the ground crew can send rescuers to that area. This saves time and resources in that it can search a wider area quicker than traditional methods and it is not necessary to send in humans until the location of the lost hikers have been
In land surveying, a number of conventional devices have been used in producing terrain mapping particularly DTM and DSM. There are such as total station , global positioning system (GPS) , light detection ranging radar (LiDAR) [3-4], manned aircraft [5-6], terrestrial laser scanning (TLS)  and remote sensing [8-9]. However, despite have been benefitted many, these approaches suffer from certain limitations particularly in terms of time consumption, usage and costing. The issue is much more serious in the tropical regions which are known persistently covered with clouds especially during monsoon seasons, making it difficult to capture high-quality images even by using remote sensing satellite technology. Meanwhile, GPS survey requires a lot of time to establish high-density points in the study area. This is because GPS survey method measures discrete point on the surface. Therefore, this method is not practical for projects allocated with limited budget and time . Terrain mapping using LiDAR and manned aircraft are very costly but has low ground resolution and limited time frame hence, rather impractical to be used for low altitude and small area surveying. Recently, UAV has been given a great attention in many applications including terrestrial terrain mapping, mainly, due to its low cost and practicality [10-11]. A UAV is commonly integrated with autopilot technology that enables semi or full autonomous navigation and image acquisition capabilities . The image acquisition capabilities enable Earth terrain to be mapped and modelled to produce orthophoto. Orthophoto is an aerial photograph that has been geometrically rectified with appropriate scale and curvature, which has been considered as a vital element in the field of photogrammetry. Besides orthophoto, images acquired from UAV can also be used to generate Digital Terrain Model (DTM), which is the spatial terrain elevations of bare-earth, DTM can be utilized
Th e innovation of this new type of unmanned aircraft product predominantly consists in that it integrates the features of a rotorcraft and a ﬁ xed-wing aircraft. Th e desired features of the multi-rotor vehicle include: vertical take-oﬀ , hovering during ﬂ ight and safe landing in virtually any conditions. However, with regard to drawbacks, this UAV shows high demand for energy, which furthermore can only be stored in batteries, which in turn means the vehicle has a limited range. Fixed- wing aircraft are free from such limitations: they combine an extensive range with considerably higher cruising speeds. An aircraft that would integrate these features would become a perfect solution for monitoring gas supply networks. Currently, the inspection is conducted by highly specialised pilots who, working in teams with an observer, perform ﬂ ight missions over gas supply networks in a conventional aircraft. Furthermore, gas leakage is most commonly diagnosed organoleptically, e.g. on the basis of a change in the colour of vegetation in the area of the spill – since methane causes it to acquire a yellow colour. Th erefore, in Polish conditions, the detection is limited to the spring/summer period, owing to the snow cover. In winter scenarios, when the ground is frozen, the hazard from the severed gas pipeline may actually occur at a distance from the conduit. Th e frozen soil does not allow gas to escape into the atmosphere; therefore, it may travel through underground channels and ﬁ nd outlets where strict anti-explosive regulations do not apply (Pożar gazociągu… 2018).
The simulation scenario requires the UAV to reduce the airspeed from 35m/s to 25m/s while maintaining a constant height at 100m. Since the airspeed spans a large range across the flight envelope, the nonlinearity of the dynamics will be excited. The tracking results of this case study are given in Fig.7 and 8, which manifest that the controller design based on the NDI technique is able to deal with the nonlinear dynamics and provide stability. However, the baseline NDI controller in this scenario demonstrates notable tracking errors especially on the height channel due to the combination of the external disturbance and internal uncertainties. The NDI controller with integral action gives a better tracking accuracy than the baseline controller. Nevertheless, compared to DOBC it yields a more oscillatory outputs and its correction is much slower as it needs the error signals to accumulate adequately to take actions. This becomes more obvious at 60s of the simulation when the vertical wind speed starts to change from −2m/s to 2m/s. The corresponding height output under the integral NDI control results in a large deviation from the reference signal, whereas the output under DOBC quickly recovers to the reference signal because it exploits the estimated disturbances in a feed-forward fashion.
produce an inexpensive and accurate property map that could be used for the adjudication of land borders, geographical precision was a very important consideration. To that end, the researchers designated and surveyed a total of 23 ground control points throughout the area to be mapped. The numerous ground control points were used to ensure that some would remain if the markers were blown away by the wind or removed by local people. The mission was a success, producing a map that could be used for enforcing customary land rights boundaries. In some cases, GPS receivers and IMUs (inertial measurement units) whose intended use is navigation and control are accurate enough to produce usable results for mapping. However, many simple drones do not log their GPS coordinates, but merely use the onboard GPS to feed data into the autopilot system. GPS loggers, such as the Flytrex Core 2 Flight Tracker, collect longitude, latitude, and altitude values during flight, using the same GPS chip used for navigation, in data formats that can be used to help geo reference maps. Some digital cameras, such as the Canon S100, come with the ability to track the GPS location of where each image was captured, producing data that can then be used to geo reference the image with processing software—although the positional accuracy is not as high as that obtained with ground control points.
The Dragan Flyer X4-ES is an UnmannedAerial Vehicle (UAV) equipped with a Sony high-definition RX100-III 20.1mp camera and LiPo long-lasting high power batteries. This drone equipped with a powerful high-resolution camera, and a Carl Zeiss zoom lens mounted on a 2-axis stability gimbal this drone technology by the company Dranganfly, is potentially a market-ready innovation providing eyes-in the-sky at a fraction of the operating cost. With a Handheld Ground Control System (GCS) video is broadcast to a receiver in control of flight operations. Both still imagery documentation capabilities, as well as video documentation make this drone a useful tool for various commercial applications. Equipped with an onboard computer and 11 sensors the helicopter has a park mode, or GPS holding position, ensure an easy to use piloting critical for real-time video and photos. Flying the drone requires a two-day authorized flight-training seminar ($2300) where upon completion an operator can become familiar with basic flight drills, payload stability and flight safety in certain weather conditions. At a purchase price of $43,225.27 Taxes included, the drone and inspection equipment are together perhaps just an expensive hobby toy without a commercial market application to provide a suitable Return on this Asset (ROA). Again the purpose of this document is to determine the value proposition of this specific drone relative to various market applications; this is the point where technology intersects markets.
Framework for software exploits: This idea relates to the discarded research questions, as it has been considered as the main topic for investigation before the focus was changed to the current research questions. Commercial UAVs are mostly pre-programmed and not steered by hand any longer. A remote control is usually available as a backup, but is not used in most cases, as even start and landing can be performed fully autonomously. This makes the software used to program the flight computer a valuable target. If a person were able to install malware on the device used for programming the flight computer, full control might be gained over the UAV by the attacker. However, there are multiple flight planning software vendors. An ap- proach could be to develop malware targeted on modifying flight information for one vendor (Open Source preferably, as the code can be reviewed) and then gradually extending this Proof-of-Concept to a framework containing multiple software vendors. Once this level is reached the malware could be used to compromise several UAVs of different manufacturers. A disadvantage of this idea is that it does not provide a solution to an intruding UAV which was not previously targeted by the malware. The malware has always to be installed be- fore in order to be able to take over control.