As the drone market expands, the corresponding standardization follows. Drone standardization can vary geographically based on the regulations and requirements of different areas. This study mainly focuses on the European Union Aviation Safety Agency (EASA) regulations and investigates Aerit’s role, as a drone operator in Sweden, within this standardization framework. In particular, Specific Operations and Risk Assessment (SORA) process, developed by EASA, is illustrated. The process covers a comprehensive range of factors related to drone operations to assess and manage risks. In addition to the drone design standardization process, the study looks into drone flight control systems at component-level redundancy and at system-level redundancy with a scientific grounding of dependability. An investigation of what a voting system looks like is then conducted for implementing a redundant flight control architecture. Furthermore, results from Software-In-The-Loop (SITL) implementation in this study show that the performance differs not much for the two flight control architectures (component-level and system-level). Thus, the decision of whether to use one flight controller or redundant flight controllers depends on the specific requirements and priorities of the drone application as well as the level of pre-flight testing. / Eftersom drönarmarknaden växer, följer motsvarande standardisering med. Standardisering av drönar kan variera geografiskt baserat på olika områdens lagar och krav. Denna studie fokuserar främst på Europeiska unionens byrå för luftfartssäkerhet (EASA) och undersöker Aerits roll som drönaroperatör i Sverige inom detta standardiseringsramverk. Särskilt beskrivs processen för Specifika Operationer och Riskbedömning (SORA), utvecklad av EASA. Denna process täcker ett omfattande utbud av faktorer relaterade till drönaroperationer för att bedöma och hantera risker. Utöver standardiseringsprocessen för drönardesign, granskar studien drönarflygkontrollsystem på komponentnivå för redundans och på systemnivå med en vetenskaplig grund för tillförlitlighet. En undersökning av hur ett röstningssystem ser ut genomförs sedan för att implementera en redundant flygkontrollarkitektur. Vidare visar resultaten från mjukvara-i-slingan (SITL) -implementeringen i denna studie att prestandan inte skiljer sig mycket mellan de två flygkontrollarkitekturerna (komponentnivå och systemnivå). Därför beror beslutet om att använda en flygkontroll eller redundanta flygkontroller på de specifika kraven och prioriteterna för drönapplikationen samt nivån av före-flygtestning.

This thesis presents the development of self adaptive flight controller for an unmanned helicopter system under hovering manoeuvre. The neural network (NN) based model predictive control (MPC) approach is utilised in this work. We use this controller due to its ability to handle system constraints and the time varying nature of the helicopter dynamics. The non-linear NN based MPC controller is known to produce slow solution convergence due to high computation demand in the optimisation process. To solve this problem, the automatic flight controller system is designed using the NN based approximate predictive control (NNAPC) approach that relies on extraction of linear models from the non-linear NN model at each time step. The sequence of control input is generated using the prediction from the linearised model and the optimisation routine of MPC subject to the imposed hard constraints. In this project, the optimisation of the MPC objective criterion is implemented using simple and fast computation of the Hildreth's Quadratic Programming (QP) procedure. The system identification of the helicopter dynamics is typically performed using the time regression network (NNARX) with the input variables. Their time lags are fed into a static feed-forward network such as the multi-layered perceptron (MLP) network. NN based modelling that uses the NNARX structure to represent a dynamical system usually requires a priori knowledge about the model order of the system. Low model order assumption generally leads to deterioration of model prediction accuracy. Furthermore, massive amount of weights in the standard NNARX model can result in an increased NN training time and limit the application of the NNARX model in a real-time application. In this thesis, three types of NN architectures are considered to represent the time regression network: the multi-layered perceptron (MLP), the hybrid multi-layered perceptron (HMLP) and the modified Elman network. The latter two architectures are introduced to improve the training time and the convergence rate of the NN model. The model structures for the proposed architecture are selected using the proposed Lipschitz coefficient and k-cross validation methods to determine the best network configuration that guarantees good generalisation performance for model prediction. Most NN based modelling techniques attempt to model the time varying dynamics of a helicopter system using the off-line modelling approach which are incapable of representing the entire operating points of the flight envelope very well. Past research works attempt to update the NN model during flight using the mini-batch Levenberg-Marquardt (LM) training. However, due to the limited processing power available in the real-time processor, such approaches can only be employed to relatively small networks and they are limited to model uncoupled helicopter dynamics. In order to accommodate the time-varying properties of helicopter dynamics which change frequently during flight, a recursive Gauss-Newton (rGN) algorithm is developed to properly track the dynamics of the system under consideration. It is found that the predicted response from the off-line trained neural network model is suitable for modelling the UAS helicopter dynamics correctly. The model structure of the MLP network can be identified correctly using the proposed validation methods. Further comparison with model structure selection from previous studies shows that the identified model structure using the proposed validation methods offers improvements in terms of generalisation error. Moreover, the minimum number of neurons to be included in the model can be easily determined using the proposed cross validation method. The HMLP and modified Elman networks are proposed in this work to reduce the total number of weights used in the standard MLP network. Reduction in the total number of weights in the network structure contributes significantly to the reduction in the computation time needed to train the NN model. Based on the validation test results, the model structure of the HMLP and modified Elman networks are found to be much smaller than the standard MLP network. Although the total number of weights for both of the HMLP and modified Elman networks are lower than the MLP network, the prediction performance of both of the NN models are on par with the prediction quality of the MLP network. The identification results further indicate that the rGN algorithm is more adaptive to the changes in dynamic properties, although the generalisation error of repeated rGN is slightly higher than the off-line LM method. The rGN method is found capable of producing satisfactory prediction accuracy even though the model structure is not accurately defined. The recursive method presented here in this work is suitable to model the UAS helicopter in real time within the control sampling time and computational resource constraints. Moreover, the implementation of proposed network architectures such as the HMLP and modified Elman networks is found to improve the learning rate of NN prediction. These positive findings inspire the implementation of the real time recursive learning of NN models for the proposed MPC controller. The proposed system identification and hovering control of the unmanned helicopter system are validated in a 6 degree of freedom (DOF) safety test rig. The experimental results confirm the effectiveness and the robustness of the proposed controller under disturbances and parameter changes of the dynamic system.

No description available.

No description available.

Testning av flygkontroller på en verklig kvadrokopter är en farlig och utmanande process eftersom kvadrokoptern kan krascha om flygkontrollern inte fungerar eller ifall operatören saknar flygerfarenheter. Den nuvarande lösningen är att montera kvadrokoptern i en mekanisk anordning som möjliggör testning av flygkontroller i säkra miljöer. Problemet med de befintliga testanordningarna är att de inte kan ge realistiska simuleringsförhållanden eftersom de i viss mån påverkar kvadrokopterns rörelse med följden att flygkontroller som utvecklas har begränsad grad av kontroll på kvadrokoptern. Syftet är att utforma en mekanisk anordning som ska ge möjlighet att både testa samt övningsflyga kvadrokoptern utan risk för personliga och materiella skador. Målet är att ta fram ritningar som ska kunna utgöra underlag för framtagning av en testprototyp. En litteraturstudie på befintliga testanordningar genomfördes som användes som underlag tillsammans med kvalitetshuset för att ta fram en kravspecifikation. Därefter genererades ett antal koncept som utvärderades med hjälp av beslutsmatris. Det valda konceptet modellerades sedan i CAD-programmet och utifrån den virtuella modellen konstruerades en verklig modell i trä som testades för att verifiera dess funktion. Resultatet är en fjäderbalanserad testanordning med sex frihetsgrader. Det är en konstruktion i aluminium innehållande en mekanisk arm som ger tre translationsfrihetsgrader, ett kulledsfäste som ger tre rotationsfrihetsgrader samt dragfjädrar för att tyngdkraftskompensera systemet. Testning av trä-modellen uppvisar att kvadrakoptern måste framföras i full fart för att styras tillsammans med armen eftersom friktionen mellan testanordningens leder är hög. Under förutsättning att friktionen mellan lederna kan hanteras verkar det att testanordningen uppfyller de ställda teoretiska förutsättningarna för att inte ha någon väsentlig påverkan på kvadrokoptern. Ändå kravs det kvalificerade tester innan något påstående att testanordningen inte påverkar kvadrokopterns rörelse och därmed kan ge realistiska flygsimuleringsförhållanden, ska kunna anges. / Testing of autonomous flight controllers on a real quadrocopter is a dangerous and challenging process because the quadrocopter can crash in case the flight controller does not function properly or in case the operator has no flight experience. The current solution is to mount the quadrocopter on a teststand, which allow the testing of flight controller in safe environments. The problem with the existing teststands is that they cannot provide realistic free flight conditions as they, to some extent, affect quadrocopter’s movement. Consequently, the developed flight controller is partially able to control the quadrocopter. The purpose with this study is to design a mechanical device for use in testing and learning to fly a quadrocopter without the risk of crashing the flying model or harming the people involved. The goal is to provide drawings for developing a test prototype.In order to understand the problem a literature review of previous test devices was carried out. The findings from the literature review were used in combination with Quality Function Deployment technique to create a House of Quality and thus develop a set of engineering specifications. After that, a number of concepts was generated and then evaluated by Pugh’s method. The selected concept was modeled in the CAD-software and based on the virtual model, a real model made of wood was constructed and tested in order to verify the function of the testbed. The final result is a spring-balanced test device with six degrees of freedom. It is a structure consisting of a mechanical arm providing three translational degrees of freedom, a swivel joint with three rotational degrees of freedom and a set of extension springs to achieve gravity balancing. The experimental results from the wooden model shows that the quadrocopter is required to fly at full speed in order for it to operate with the arm due to the high friction between the joints. Under the condition that the friction between the joints can be managed, the test device seems to fulfill the theoretical requirements for simulating free flight condition. Nevertheless, it requires specialized and advanced testing before any assertion that the test device does not affect the dynamics of quadrocopter and thus it can provide completely realistic flight conditions, can be made.

AirSim is a new simulator developed as a plugin for the Unreal Engine, aiming to be a useful tool aiding the development and testing of algorithms for autonomous vehicles. Due to AirSim still being in its infancy there is little to none research available of its possibilities or detailed guidelines and tutorials on how to use its APIs.Rust is a new systems programming language with the purpose of being safe, practical and concurrent which through design choices can solve some of the major drawbacks of the C programming language such as memory leaks, thread management, and segmentation faults.Researching the features of AirSim and its different ways of communicating, we determine the possibility of implementing a custom flight controller in Rust and C able to control a drone in the simulator and evaluate the capabilities of Rust compared to C. This is conducted by reading available documentation for AirSim, studying the source code and learning about the communication protocols used by AirSim.This thesis results in an implementation of a custom flight controller in Rust and C that controls a drone in AirSim using a communication protocol named MAVLink which enables fine-grained control of the motors. The conclusion made about the comparison of Rust and C is that both languages were able to implement the safety-critical functionality of the flight controller and that Rust provided capabilities which could be useful when developing safety-critical systems. / AirSim är en ny simulator utvecklad som ett plugin för Unreal Engine, med målet att fungera som ett hjälpmedel inom utveckling och testning av algoritmer för autonoma fordon. På grund av att AirSim fortfarande är väldigt ungt finns väldigt lite forskning tillgänglig om dess möjligheter eller detaljerade riktlinjer och beskrivningar för användningen av dess APIer.Rust är ett nytt programmeringsspråk med målet att vara säkert, praktiskt och parallellt vilket genom designval kan lösa några av de största problemen med programmeringsspråket C som till exempel minnessläckor, trådhantering och segmenteringsfel.Genom att undersöka funktionerna i AirSim och dess olika sätt att kommunicera, utforskar vi möjligheten av att utveckla en egen flygkontroller i Rust och C som kan styra en drönare i simulatorn och utvärdera Rust i förhållande till C. Detta genomförs genom att läsa tillgänglig dokumentation för AirSim, studera källkoden och lära oss de kommunikationsprotokoll som används av AirSim.Denna avhandling resulterar i implementationen av en egen flygkontroller i Rust och C som styr en drönare i AirSim med kommunikationsprotokollet MAVLink, vilket möjliggör en noggrann kontroll av motorerna. Slutsatsen gällande Rust och C är att båda språken fungerade väl för implementationen av säkerhetsritiska funktioner i flygkontrollern samt att Rust erbjöd förmågor som kan visa sig vara användbara vid utveckling av säkerhetskritiska system.

In recent times the unmanned quadcopter aircraft has been used for a widening range of applications, but for longer distances it still falls short of conventional airplanes in terms of energy usage. There exists hybrid configurations of unmanned aircraft which combine the mobility of quadcopters and the range of fixed-wing aircraft. The transition between the hovering mode and the gliding mode, however, is a complex non-linear control problem. To solve this a recent study applied a neural network as a closed loop controller. This controller was capable of seamless mode transition and could be trained for any copter configuration using reinforcement learning. The work presented here focuses on improvements to the method of controller design established by said study, mainly focusing on increased realism of the aerodynamic simulations and the addition of control surfaces for increased maneuverability. These improvements resulted in a lift of 37% of the total copter weight and a higher achievable top speed of 8 m/s before instability occurs. To verify these improvements were not only present in the simulations a physical prototype was also constructed which when tested succeeded in hovering flight but failed to sustain any significant forward flight. / På senare tid har obemannade quadcopters kraftigt expanderat sina användningsområden, men för längre sträckor slås de fortfarande av konventionella flygplan när det gäller energiåtgång. Det finns hybridkonfigurationer av obemannade farkoster som kombinerar quadcopterns rörlighet och räckvidden av flygplan. Övergången mellan hovrande läge och horisontell flygning är emellertid ett komplext icke-linjärt reglerproblem. För att lösa detta använde en nyligen genomförd studie ett neuralt nätverk som en regulator i ett återkopplat system. Den här styrenheten kunde sömlöst övergå mellan flyglägen och kunde tränas för valfri copterkonfiguration med hjälp av reinforcement learning. Arbetet som presenteras här fokuserar på förbättringar av metoden för regulatordesign som fastställts av nämnda studie, främst med fokus på ökad realism av de aerodynamiska simuleringarna och tillägget av kontrollytor för ökad manövrerbarhet. Dessa förbättringar resulterade i en genererad lyftkraft upp till 37% av farkostens vikt och en förhöjd maxhastighet till 8 m/s före instabilitet. För att verifiera dessa resultat i verkligheten konstruerades en fysisk prototyp som vid försök lyckades stabilisera sig i hovrande läge men inte upprätthålla någon signifikant framåtfart.