The manufacturing of connected and automated vehicles (CAVs) is happening and they are aiming at providing an efficient, safe, and seamless driving experience. This is done by offering automated driving together with wireless communication to and from various objects in the surrounding environment. How automated the vehicle is can be classified from level 0 (no automation at all) to level 5 (fully automated). There is many potential attack vectors of CAVs for attackers to take advantage of and these attack vectors may change depending on what level of automation the vehicle have. There are some known vulnerabilities of CAVs where the security has been breached, but what is seemed to be lacking in the academia in the field of CAVs is a place where the majority of information regarding known attack vectors and cyber-attacks on those is collected. In addition to this the attack vectors may be analyzed for each level of automation the vehicles may have. This research is a systematic literature review (SLR) with three stages (planning, conducting, and report) based on literature review methodology presented by Kitchenham (2004). These stages aim at planning the review, finding articles, extracting information from the found articles, and finally analyzing the result of them. The literature review resulted in information regarding identified cyberattacks and attack vectors the attackers may use as a path to exploit vulnerabilities of a CAV. In total 24 types of attack vectors were identified. Some attack vectors like vehicle communication types, vehicle applications, CAN bus protocol, and broadcasted messages were highlighted the most by the authors. When the attack vectors were analyzed together with the standard of ‘Levels of Driving Automation’ it became clear that there are more vulnerabilities to consider the higher level of automation the vehicle have. The contributions of this research are hence (1) a broad summary of attack vectors of CAVs and (2) a summary of these attack vectors for every level of driving automation. This had not been done before and was found to be lacking in the academia.

Platooning has the potential to be a breakthrough in increasing road capacity and reducing fuel consumption, as it allows a chain of vehicles to closely follow each other on the road. When the number of vehicles increases, platoons will follow one another in what is referred to as multiplatooning. Many Cooperative Intelligent Transportation Systems (C-ITS) applications rely on periodically exchanged beacons among vehicles to improve traffic safety. However, as the number of connected vehicles increases, the network may become congested due to periodically exchanged beacons. Therefore, without some congestion control method, safety critical messages such as Cooperative Awareness Messages (CAMs) may not be delivered on time in high vehicle density scenarios. Both the European Telecommunications Standards Institute (ETSI) and the Institute of Electrical and Electronics Engineers (IEEE) have been working on different standards to support vehicular communication. ETSI dened the Decentralized Congestion Control (DCC) mechanism which adapts transmission parameters (message rate, transmit data rate, and transmit power, etc.) to keep channel load under control. ETSI DCC utilizes a three-state machine with RELAXED, ACTIVE, and RESTRICTIVE states. In this thesis, we implemented this three-state machine by adapting the message rate based on the channel busy ratio (CBR). We name this message-rate based three-state machine DCC-3. DCC-3 has the ability to control channel load; however, it has unfairness and instability problems due to the dramatic parameter changes between states. Therefore, we divided the ACTIVE state of DCC-3 into ve sub-states, and refer to this as DCC-7. We benchmarked DCC-3 against static beaconing (STB), dynamic beaconing (DynB), LInear MEssage Rate Integrated Control (LIMERIC), and DCC-7 using different evaluation metrics with different numbers of platoons. Our results from the Plexe simulator demonstrate that DCC-7 has the best performance when considering all evaluation metrics, including CBR, Inter-reception time (IRT), collisions, safe time ratio, and fairness. Furthermore, we found using transmit power control could greatly improve the performance of CBR and collision rates. / Platooning (fordonskonvojer) har potential att bli ett genombrott i öka vägkapaciteten och minska bränsleförbrukning, eftersom det tillåter en kedja av fordon att noga följa varandra på vägen. När antalet fordon ökar, kommer att plutoner följa varandra i vad som benämns multiplatooning (konvoj av fordonskonvojer). Många kooperativ intelligenta transportsystem (C-ITS) tillämpningar förlitar sig på regelbundet utbytte beacons bland fordon att förbättra traffiksäkerheten. Dock som antalet uppkopplade fordon ökar, kan nätverket bli överbelastat på grund av regelbundet utbytte beacons. Utan någon trängsel kontrollmetod, får därför säkerhet kritiska meddelanden såsom kooperativ medvetenhet meddelanden (CAMs) inte levereras i tid i höga fordon densitet scenarier. Både Europeiska institutet för telekommunikationsstandarder (ETSI) och Institute el och elektroniska tekniker (IEEE) har arbetat på olika standarder för att stödja vehicular kommunikation. ETSI definieras den decentraliserade överbelastning kontroll (DCC) mekanism som anpassar överföring parametrar (meddelande hastighet, överföra datahastighet och sändningseffekt, etc.) för att hålla kanalen belastningen under kontroll. ETSI DCC använder en tre-state maskin med RELAXED, ACTIVE och RESTRICTIVE stater. I denna avhandling har genomfört vi denna tre-state maskin genom att anpassa meddelande hastighet baserat på kanal upptagen förhållandet (CBR). Vi nämna detta meddelande-hastighet baserat tre-state machine DCC-3. DCC-3 har förmågan att kontrollera kanal belastning; Det har dock otillbörlighet och instabilitet problem på grund av de dramatiska parameterändringar mellan stater. Därför vi indelat det ACTIVE tillståndet för DCC-3 i fem undertillstånd och hänvisar till detta som DCC-7. Vi benchmarkade DCC-3 mot statiska leda (STB), dynamisk leda (DynB), linjära MEssage Rate integrerad kontroll (LIMERIC) och DCC-7 med olika utvärdering statistik med olika antal plutoner. Våra resultat från Plexe simulator visar att DCC-7 har bästa prestanda när man överväger alla utvärdering statistik, inklusive CBR, mellan receptionen tid (IRT), kollisioner, säker tid baserat och rättvisa. Vi fann dessutom använda Sändareffektstyrning kan avsevärt förbättra prestanda för CBR och kollision priser.

Rede Ad hoc Veicular (VANET) é um subconjunto singular das redes ad hoc móveis (MANET), com o diferencial de que os nós são veículos providos de tecnologia própria de comunicação e que interagem para formar redes espontâneas, valendo-se de pouca ou nenhuma infraestrutura estabelecida previamente. VANETs admitem a integração de diferentes tecnologias sem fio na pretensão de mitigar adversidades, agregar segurança e eficiência ao tráfego. Na disseminação de dados, um salto único é suficiente para orientar os elementos ao alcance do sinal de rádio, e nós intermediários sustentam a comunicação aos demais, em encaminhamento multihop. Amparados em dispositivos embarcados, os veículos produzem registros, detectam sinais, trocam advertências e métricas. Avaliações dessas informações permitem ao condutor decisões ou reações antecipadas em situações adversas, a exemplo dos acidentes ou congestionamentos. Nesse contexto, a execução deste trabalho trata questões para elaborar estratégias adaptativas inteligentes de disseminação de dados, uma vez que as mesmas se consolidam como lastros da comunicação em VANET com condições adversas de operação. A abordagem proposta se utiliza de sistemas fuzzy para a detecção de congestionamento, com o propósito de agregar autonomia e adaptar a estratégia de disseminação às condições de tráfego identificadas. A convergência nos desenvolvimentos realizados se reflete na estratégia eFIRST, uma solução robusta para a detecção autônoma da condição atual de congestionamento que resguarda a disseminação adaptativa de alertas e abranda o problema da interrupção no tráfego. A abordagem se sustenta apenas na comunicação entre veículos e nos registros de identificação da vizinhança local, agregados em uma estratégia fuzzy e no ajuste adaptativo da potência do sinal de transmissão. Em conformidade com as tendências de condução e com os sistemas inteligentes, este desenvolvimento contribui com subsídios para ratificar a aproximação fuzzy como estratégia adaptativa às flutuações na densidade veicular, em diferentes cenários e regimes de tráfego. As avaliações comparativas do eFIRST respaldam concluir que a estratégia oportuniza o equilíbrio otimizado das perdas, colisões e cobertura, com superior alcance de propagação e redução dos congestionamentos. / Vehicular Ad hoc Network (VANET) is a unique subset of mobile ad hoc networks (MANET), with the difference that nodes are vehicles provided with their own communication technology and interact to form spontaneous networks, with little or no infrastructure previously established. VANETs support the integration of different wireless technologies in order to mitigate adversities, add security and efficiency to traffic. In the data dissemination, a single hop is sufficient to guide the elements within reach of the radio signal, and intermediary nodes support the communication with the others in multihop routing. Supported by embedded devices, vehicles produce records, detect signals, exchange warnings and metrics. Assessments of this information allow the driver to make decisions or react beforehand in adverse situations, such as accidents or traffic congestions. From the observations in this context, this work deals with questions to elaborate intelligent adaptive strategies in data dissemination, since they consolidate themselves as ballast communication in VANET with adverse operating conditions. The proposed approach uses fuzzy systems to detect traffic congestion, with the purpose of aggregating autonomy and adapting the dissemination strategy to the identified traffic conditions. The convergence in the developments performed is reflected in the eFIRST strategy, a robust solution for the autonomous detection of the current traffic congestion condition that protects the adaptive dissemination of alerts and reduces the problem of the interruption in the traffic. The approach is supported only by communication between vehicles and in local neighborhood identification records, aggregated in a fuzzy strategy and in the adaptive adjustment of transmission signal power. In accordance with the driving trends and with the intelligent systems, this development contributes with assistance for ratify the fuzzy approach as an adaptive strategy to fluctuations in vehicular density in different scenarios and traffic regimes. Comparative evaluations of eFIRST support the conclusion that the strategy favors the optimal balance of losses, collisions and coverage, with a greater range of propagation and reduction of congestion.

[pt] A perda de percurso (PL) é um parâmetro essencial em modelos de propagação e crucial na determinação da área de cobertura de sistemas móveis. Os métodos de aprendizado de máquina (ML) tornaram-se ferramentas promissoras para a previsão de propagação de rádio. No entanto, ainda existem alguns desafios para sua implantação completa, relacionados à seleção das entradas mais significativas do modelo, à compreensão de suas contribuições para as previsões do modelo e à avaliação adicional da capacidade de generalização para amostras desconhecidas. Esta tese tem como objetivo projetar modelos de PL baseados em ML otimizados para diferentes aplicações das tecnologias 5G e além. Essas aplicações abrangem links de ondas milimétricas (mmWave) para ambientes indoor e outdoor na faixa de frequência de 26,5 a 40 GHz, cobertura de macrocélulas no espectro sub-6 GHz e comunicações veiculares usando campanhas de medições desenvolvidas em CETUC, Rio de Janeiro, Brazil. Vários algoritmos de ML são explorados, como redes neurais artificiais (ANN), regressão de vetor de suporte (SVR), floresta aleatória (RF) e aumento de árvore de gradiente (GTB). Além disso, estendemos dois modelos empíricos para mmWave com previsão de PL melhorada. Propomos uma metodologia para seleção robusta de modelos de ML e uma metodologia para selecionar os preditores mais adequados para as máquinas consideradas com base na melhoria de desempenho e na interpretabilidade do modelo. Além disso, para o canal veículo-veículo (V2V), uma técnica de rede neural convolucional (CNN) também é proposta usando uma abordagem de aprendizado por transferência para lidar com conjuntos de dados pequenos. Os testes de generalização propostos mostram a capacidade dos modelos de ML de aprender o padrão entre as entradas do modelo e a PL, mesmo em ambientes e cenários mais desafiadores de amostras desconhecidas. / [en] Path loss (PL) is an essential parameter in propagation models and critical in determining mobile systems’ coverage area. Machine learning (ML) methods have become promising tools for radio propagation prediction. However, there are still some challenges for its full deployment, concerning to selection of the most significant model s inputs, understanding their contributions to the model s predictions, and a further evaluation of the generalization capacity for unknown samples. This thesis aims to design optimized ML-based PL models for different applications for the 5G and beyond technologies. These applications encompass millimeter wave (mmWave) links for indoor and outdoor environments in the frequency band from 26.5 to 40 GHz, macrocell coverage in the sub-6 GHz spectrum, and vehicular communications using measurements campaign carried out by the Laboratory of Radio-propagation, CETUC, in Rio de Janeiro, Brazil. Several ML algorithms are exploited, such as artificial neural network (ANN), support vector regression (SVR), random forest (RF), and gradient tree boosting (GTB). Furthermore, we have extended two empirical models for mmWave with improved PL prediction. We proposes a methodology for robust ML model selection and a methodology to select the most suitable predictors for the machines considered based on performance improvement and the model’s interpretability. In adittion, for the vehicle-to-vehicle (V2V) channel, a convolutional neural network (CNN) technique is also proposed using a transfer learning approach to deal with small datasets. The generalization tests proposed shows the ability of the ML models to learn the pattern between the model’s inputs and PL, even in more challenging environments and scenarios of unknown samples.

[pt] A perda de percurso (PL) é um parâmetro essencial em modelos de propagação e crucial na determinação da área de cobertura de sistemas móveis. Osmétodos de aprendizado de máquina (ML) tornaram-se ferramentas promissoras para a previsão de propagação de rádio. No entanto, ainda existem algunsdesafios para sua implantação completa, relacionados à seleção das entradasmais significativas do modelo, à compreensão de suas contribuições para asprevisões do modelo e à avaliação adicional da capacidade de generalizaçãopara amostras desconhecidas. Esta tese tem como objetivo projetar modelosde PL baseados em ML otimizados para diferentes aplicações das tecnologias5G e além. Essas aplicações abrangem links de ondas milimétricas (mmWave)para ambientes indoor e outdoor na faixa de frequência de 26,5 a 40 GHz,cobertura de macrocélulas no espectro sub-6 GHz e comunicações veicularesusando campanhas de medições desenvolvidas em CETUC, Rio de Janeiro,Brazil. Vários algoritmos de ML são explorados, como redes neurais artificiais(ANN), regressão de vetor de suporte (SVR), floresta aleatória (RF) e aumentode árvore de gradiente (GTB). Além disso, estendemos dois modelos empíricospara mmWave com previsão de PL melhorada. Propomos uma metodologiapara seleção robusta de modelos de ML e uma metodologia para selecionar ospreditores mais adequados para as máquinas consideradas com base na melhoria de desempenho e na interpretabilidade do modelo. Além disso, para o canalveículo-veículo (V2V), uma técnica de rede neural convolucional (CNN) também é proposta usando uma abordagem de aprendizado por transferência paralidar com conjuntos de dados pequenos. Os testes de generalização propostosmostram a capacidade dos modelos de ML de aprender o padrão entre as entradas do modelo e a PL, mesmo em ambientes e cenários mais desafiadoresde amostras desconhecidas. / [en] Path loss (PL) is an essential parameter in propagation models and critical in determining mobile systems coverage area. Machine learning (ML) methods have become promising tools for radio propagation prediction. However, there are still some challenges for its full deployment, concerning to selection of the most significant model s inputs, understanding their contributions to the model s predictions, and a further evaluation of the generalization capacity for unknown samples. This thesis aims to design optimized ML-based PL models for different applications for the 5G and beyond technologies. These applications encompass millimeter wave (mmWave) links for indoor and outdoor environments in the frequency band from 26.5 to 40 GHz, macrocell coverage in the sub-6 GHz spectrum, and vehicular communications using measurements campaign carried out by the Laboratory of Radio-propagation, CETUC, in Rio de Janeiro, Brazil. Several ML algorithms are exploited, such as artificial neural network (ANN), support vector regression (SVR), random forest (RF), and gradient tree boosting (GTB). Furthermore, we have extended two empirical models for mmWave with improved PL prediction. We proposes a methodology for robust ML model selection and a methodology to select the most suitable predictors for the machines considered based on performance improvement and the model s interpretability. In adittion, for the vehicle-to-vehicle (V2V) channel, a convolutional neural network (CNN) technique is also proposed using a transfer learning approach to deal with small datasets. The generalization tests proposed shows the ability of the ML models to learn the pattern between the model’s inputs and PL, even in more challenging environments and scenarios of unknown samples.

No description available.

Connected and autonomous vehicles are emerging technologies that have fostered the Intelligent Transportation System (ITS). ITS has the objective of optimizing traffic safety, mobility, and fuel consumption. To achieve this, a range of different services are provided that utilize communication in a vehicular network. One of these services that has received a lot of attention lately due to its ongoing standardization is the Maneuver Coordination Service (MCS). MCS has already shown great potential in the support of complex traffic areas, also called Transition Area (TA), where vehicles must cooperate to avoid Transition of Controls (ToCs). ITS-services often rely on communicated data; small errors, such as inaccessible or incorrect data, can cause the system to behave incorrectly. Signal interference (jamming) can cause communication interruptions, making vehicles unaware of each other. Incorrect data can be intentional due to data injection attacks, but also unintentional due to malfunctioning sensors, making vehicles incorrectly aware of each other. Incorrect behavior in systems such as ITS can lead to traffic congestion or even life-threatening collisions. This study focuses on MCS and examines traffic behavior when the service, in a generic traffic scenario, is subjected to jamming and falsification attacks with a variety of strategies (negative and positive speed, acceleration and position offset). We considered external attackers (not authenticated) that can disrupt communication, as well as internal attackers (authenticated) that are limited to tampering with outgoing data. Through severe collisions and travel time delays, the results show an impact on both safety and mobility. The results also show that different attacks with different impacts on the adversary can cause similar effects on the traffic, thus allowing the adversary to choose attacks based on the desired impact and its rationality, i.e. its willingness to be part of the impact. The study also proposes an extension to an already proposed Maneuver Coordination Protocol (MCP). We show that our extended MCP can be beneficial in avoiding dangerous maneuvers that could lead to collisions with cars in the blind spot. / Uppkopplade och autonoma fordon är framväxande teknologier som har främjat Intelligenta Transporteringssystem (ITS). ITS har som mål att optimera trafiksäkerhet, mobilitet och bränsleförbrukning. För att uppnå detta tillhandahålls en rad olika tjänster som utnyttjar kommunikation i ett fordonsnät. En av dessa tjänster som har fått mycket uppmärksamhet under den senaste tiden, tack vare sin pågående standardisering, är Manöverkoordinationtjänsten (MCS). MCS har redan visat stor potential för att stödja komplexa trafikområden, även kallade Övergångsområden (TA), där fordon måste samarbeta för att undvika kontrollövergångar (ToCs). ITS-tjänster förlitar sig ofta på kommunicerad data; små fel, som otillgängliga eller felaktiga data, kan göra att systemet beter sig felaktigt. Signalstörningar kan orsaka kommunikationsavbrott, vilket gör fordon omedvetna om varandra. Felaktig data kan vara avsiktliga på grund av datainjektionsattacker, men också oavsiktliga på grund av felaktiga sensorer, vilket gör fordon felaktigt medvetna om varandra. Felaktigt beteende i system som ITS kan leda till trafikstockningar eller till och med livshotande kollisioner. Denna studie fokuserar på MCS och undersöker trafikbeteendet när tjänsten, i ett generiskt trafikscenario, utsätts för signalstörningar och förfalskningsattacker med en mängd olika strategier (negativ och positiv hastighet, acceleration och positionsförskjutning). Vi tog hänsyn till externa angripare (ej autentiserade) som kan störa kommunikationen, såväl som interna angripare (autentiserade) som är begränsade till att manipulera utgående data. Genom allvarliga kollisioner och restidsförseningar visar resultaten en inverkan på både säkerhet och mobilitet. Resultaten visar också att olika attacker med olika inverkan på angriparen kan orsaka liknande effekter på trafiken, vilket gör att angriparen kan välja attacker baserat på den önskade effekten och rationaliteten, d.v.s. dens villighet att vara en del av påverkan. Studien föreslår också en utökning av en redan föreslagen MCP. Vi visar att vårt utökade MCP kan vara till nytta för att undvika farliga manövrar som kan leda till kollisioner med bilar i döda vinkeln.

Awareness of vehicles’ surrounding conditions is important in today’s intelligent transportation system. A wide range of effort has been put in to deploy Vehicular Communication (VC) systems to make driving conditions safer and more efficient. Vehicles are aware of their surroundings with the help of authenticated safety beacons in VC systems. Since vehicles act according to the information conveyed by such beacons, verification of beacons plays an important role in becoming aware of and predicting the status of the sender vehicle. The idea of implementing secure mechanisms to deal with a high rate of incoming beacons and processing them with high efficiency becomes a very important part of the whole VC network. The goal of this work was to implement a scheme that deals with a high rate of the incoming beacon, preserve non-repudiation of the accepted messages which contains information about the current and near-future status of the sender vehicle, and at the same time keep the computation overhead as low as possible. Along with this, maintaining user privacy from a legal point of view as well as from a technical perspective by implementing privacy-enhancing technologies. These objectives were achieved by the introduction of Timed Efficient Stream Loss-Tolerant Authentication (TESLA), periodic signature verification, and cooperative verification respectively. Four different scenarios were implemented and evaluated, starting and building upon the baseline approach. Each approach addressed the problems that were aimed at this work and results show improved scalability and efficiency with the introduction of TESLA, periodic signature verification, and cooperative verification. / Medvetenheten om fordons omgivande förhållanden är viktig i dagens intelligenta transportsystem. Ett stort antal ansträngningar har lagts ned för att distribuera VC system för att göra körförhållandena säkrare och effektivare. Fordon är medvetna om sin omgivning med hjälp av autentiserade säkerhetsfyrar i VC system. Eftersom fordon agerar enligt den information som förmedlas av sådana fyrar, spelar verifiering av fyrar en viktig roll för att bli medveten om och förutsäga avsändarfordonets status. Idén att implementera säkra mekanismer för att hantera en hög frekvens av inkommande fyrar och bearbeta dem med hög effektivitet blir en mycket viktig del av hela VC nätverket. Målet med detta arbete var att implementera ett schema som behandlar en hög hastighet för det inkommande fyren, bevara icke-förkastelse av de accepterade meddelandena som innehåller information om den aktuella och närmaste framtida statusen för avsändarfordonet och samtidigt håll beräkningen så låg som möjligt. Tillsammans med detta upprätthåller användarnas integritet ur juridisk synvinkel såväl som ur ett tekniskt perspektiv genom att implementera integritetsförbättrande teknik. Dessa mål uppnåddes genom införandet av TESLA, periodisk signatur verifiering respektive samarbets verifiering. Fyra olika scenarier implementerades och utvärderades med utgångspunkt från baslinjemetoden. Varje tillvägagångssätt tog upp de problem som riktades mot detta arbete och resultaten visar förbättrad skalbarhet och effektivitet med införandet av TESLA, periodisk signatur verifiering och samarbets verifiering.

La convergencia de las telecomunicaciones, la informática, la tecnología inalámbrica y los sistemas de transporte, va a facilitar que nuestras carreteras y autopistas nos sirvan tanto como plataforma de transporte, como de comunicaciones. Estos cambios van a revolucionar completamente cómo y cuándo vamos a acceder a determinados servicios, comunicarnos, viajar, entretenernos, y navegar, en un futuro muy cercano. Las redes vehiculares ad hoc (vehicular ad hoc networks VANETs) son redes de comunicación inalámbricas que no requieren de ningún tipo de infraestructura, y que permiten la comunicación y conducción cooperativa entre los vehículos en la carretera. Los vehículos actúan como nodos de comunicación y transmisores, formando redes dinámicas junto a otros vehículos cercanos en entornos urbanos y autopistas. Las características especiales de las redes vehiculares favorecen el desarrollo de servicios y aplicaciones atractivas y desafiantes. En esta tesis nos centramos en las aplicaciones relacionadas con la seguridad. Específicamente, desarrollamos y evaluamos un novedoso protocol que mejora la seguridad en las carreteras. Nuestra propuesta combina el uso de información de la localización de los vehículos y las características del mapa del escenario, para mejorar la diseminación de los mensajes de alerta. En las aplicaciones de seguridad para redes vehiculares, nuestra propuesta permite reducir el problema de las tormentas de difusión, mientras que se mantiene una alta efectividad en la diseminación de los mensajes hacia los vehículos cercanos. Debido a que desplegar y evaluar redes VANET supone un gran coste y una tarea dura, la metodología basada en la simulación se muestra como una metodología alternativa a la implementación real. A diferencia de otros trabajos previos, con el fin de evaluar nuestra propuesta en un entorno realista, en nuestras simulaciones tenemos muy en cuenta tanto la movilidad de los vehículos, como la transmisión de radio en entornos urbanos, especialmente cuando los edificios interfieren en la propagación de la señal de radio. Con este propósito, desarrollamos herramientas para la simulación de VANETs más precisas y realistas, mejorando tanto la modelización de la propagación de radio, como la movilidad de los vehículos, obteniendo una solución que permite integrar mapas reales en el entorno de simulación. Finalmente, evaluamos las prestaciones de nuestro protocolo propuesto haciendo uso de nuestra plataforma de simulación mejorada, evidenciando la importancia del uso de un entorno de simulación adecuado para conseguir resultados más realistas y poder obtener conclusiones más significativas. / Martínez Domínguez, FJ. (2010). Improving Vehicular ad hoc Network Protocols to Support Safety Applications in Realistic Scenarios [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/9195