京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24256号 / 情博第800号 / 新制||情||135(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 河原 達也, 教授 森 信介, 教授 伊藤 孝行 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

With the rapid development of autonomous driving and artificial intelligence technology, end-to-end autonomous driving technology has become a research hotspot. This thesis aims to explore the application of deep reinforcement learning in the realizing of end-to-end autonomous driving. We built a deep reinforcement learning virtual environment in the Carla simulator, and based on it, we trained a policy model to control a vehicle along a preplanned route. For the selection of the deep reinforcement learning algorithms, we have used the Proximal Policy Optimization algorithm due to its stable performance. Considering the complexity of end-to-end autonomous driving, we have also carefully designed a comprehensive reward function to train the policy model more efficiently. The model inputs for this study are of two types: firstly, real-time road information and vehicle state data obtained from the Carla simulator, and secondly, real-time images captured by the vehicle's front camera. In order to understand the influence of different input information on the training effect and model performance, we conducted a detailed comparative analysis. The test results showed that the accuracy and significance of the information has a significant impact on the learning effect of the agent, which in turn has a direct impact on the performance of the model. Through this study, we have not only confirmed the potential of deep reinforcement learning in the field of end-to-end autonomous driving, but also provided an important reference for future research and development of related technologies.

Free and fair elections are the basis for democracy, but conducting elections is not an easy task. Different groups of people are trying to influence the outcome of the election in their favor using the range of methods, from campaigning for a particular candidate to well-financed lobbying. Often the stakes are too high, and the methods are illegal. Two main properties of any voting scheme are the privacy of a voter’s choice and the integrity of the tally. Unfortunately, they are mutually exclusive. Integrity requires making elections transparent and auditable, but at the same time, we must preserve a voter’s privacy. It is always a trade-off between these two requirements. Current voting schemes favor privacy over auditability, and thus, they are vulnerable to voting fraud. I propose two novel voting systems that can achieve both privacy and verifiability. The first protocol is based on cryptographical primitives to ensure the integrity of the final tally and privacy of the voter. The second protocol is a simple paper-based voting scheme that achieves almost the same level of security without usage of cryptography.

No description available.

In the rapidly evolving field of autonomous driving, the integration of Deep Reinforcement Learning (DRL) promises significant advancements towards achieving reliable and efficient vehicular systems. This study presents a comprehensive examination of DRL’s application within a simulated autonomous driving context, with a focus on the nuanced impact of representation learning parameters on the performance of end-to-end models. An overview of the theoretical underpinnings of machine learning, deep learning, and reinforcement learning is provided, laying the groundwork for their application in autonomous driving scenarios. The methodology outlines a detailed framework for training autonomous vehicles in the Duckietown simulation environment, employing both non-end-to-end and end-to-end models to investigate the effectiveness of various reinforcement learning algorithms and representation learning techniques. At the heart of this research are extensive simulation experiments designed to evaluate the Proximal Policy Optimization (PPO) algorithm’s effectiveness within the established framework. The study delves into reward structures and the impact of representation learning parameters on the performance of end-to-end models. A critical comparison of the models in the validation chapter highlights the significant role of representation learning parameters in the outcomes of DRL-based autonomous driving systems. The findings reveal that meticulous adjustment of representation learning parameters markedly influences the end-to-end training process. Notably, image segmentation techniques significantly enhance feature recognizability and model performance.:Contents List of Figures List of Tables List of Abbreviations List of Symbols 1 Introduction 1.1 Autonomous Driving Overview 1.2 Problem Description 1.3 Research Structure 2 Research Background 2.1 Theoretical Basis 2.1.1 Machine Learning 2.1.2 Deep Learning 2.1.3 Reinforcement Learning 2.2 Related Work 3 Methodology 3.1 Problem Definition 3.2 Simulation Platform 3.3 Observation Space 3.3.1 Observation Space of Non-end-to-end model 3.3.2 Observation Space of end-to-end model 3.4 Action Space 3.5 Reward Shaping 3.5.1 speed penalty 3.5.2 position reward 3.6 Map and training dataset 3.6.1 Map Design 3.6.2 Training Dataset 3.7 Variational Autoencoder Structure 3.7.1 Mathematical fundation for VAE 3.8 Reinforcement Learning Framework 3.8.1 Actor-Critic Method 3.8.2 Policy Gradient 3.8.3 Trust Region Policy Optimization 3.8.4 Proximal Policy Optimization 4 Simulation Experiments 4.1 Experimental Setup 4.2 Representation Learning Model 4.3 End-to-end Model 5 Result 6 Validation and Evaluation 6.1 Validation of End-to-end Model 6.2 Evaluation of End-to-end Model 6.2.1 Comparison with Baselines 6.2.2 Comparison with Different Representation Learning Model 7 Conclusion and Future Work 7.1 Summary 7.2 Future Research

Yes / With the advent of Wireless Sensor Networks (WSN) and its immense popularity in a wide range of applications, security has been a major concern for these resource-constraint systems. Alongside security, WSNs are currently being integrated with existing technologies such as the Internet, satellite, Wi-Max, Wi-Fi, etc. in order to transmit data over long distances and hand-over network load to more powerful devices. With the focus currently being on the integration of WSNs with existing technologies, security becomes a major concern. The main security requirement for WSN-integrated networks is providing end-to-end security along with the implementation of in-processing techniques of data aggregation. This can be achieved with the implementation of Homomorphic encryption schemes which prove to be computationally inexpensive since they have considerable overheads. This paper addresses the ID-issue of the commonly used Castelluccia Mykletun Tsudik (CMT) [12] homomorphic scheme by proposing an ID slotting mechanism which carries information pertaining to the security keys responsible for the encryption of individual sensor data. The proposed scheme proves to be 93.5% lighter in terms of induced overheads and 11.86% more energy efficient along with providing efficient WSN scalability compared to the existing scheme. The paper provides analytical results comparing the proposed scheme with the existing scheme thus justifying that the modification to the existing scheme can prove highly efficient for resource-constrained WSNs.

LTE heterogeneous wireless networks promise significant increase in data rates and improved coverage through (i) the deployment of relays and cell densification, (ii) carrier aggregation to enhance bandwidth usage and (iii) by enabling nodes to have dual connectivity. These emerging cellular networks are complex and large systems which are difficult to optimize with centralized control and where mobiles need to balance spectral efficiency, power consumption and fairness constraints. In this dissertation we focus on how decentralized and autonomous mobiles in multihop cellular systems can optimize their own local objectives by taking into account end-to-end or network-wide conditions. We propose several link-adaptive schemes where nodes can adjust their transmit power, aggregate carriers and select points of access to the network (relays and/or macrocell base stations) autonomously, based on both local and global conditions. Under our approach, this is achieved by disseminating the dynamic congestion level in the backhaul links of the points of access. As nodes adapt locally, the congestion levels in the backhaul links can change, which can in turn induce them to also change their adaptation objectives. We show that under our schemes, even with this dynamic congestion feedback, nodes can distributedly converge to a stable selection of transmit power levels and points of access. We also analytically derive the transmit power levels at the equilibrium points for certain cases. Moreover, through numerical results we show that the corresponding system throughput is significantly higher than when nodes adapt greedily following traditional link layer optimization objectives. Given the growing data rate demand, increasing system complexity and the difficulty of implementing centralized cross-layer optimization frameworks, our work simplifies resource allocation in heterogeneous cellular systems. Our work can be extended to any multihop wireless system where the backhaul link capacity is limited and feedback on the dynamic congestion levels at the access points is available. / Ph. D.

Language is the principal medium for ideas, while dialogue is the most natural and effective way for humans to interact with and access information from machines. Natural language generation (NLG) is a critical component of spoken dialogue and it has a significant impact on usability and perceived quality. Many commonly used NLG systems employ rules and heuristics, which tend to generate inflexible and stylised responses without the natural variation of human language. However, the frequent repetition of identical output forms can quickly make dialogue become tedious for most real-world users. Additionally, these rules and heuristics are not scalable and hence not trivially extensible to other domains or languages. A statistical approach to language generation can learn language decisions directly from data without relying on hand-coded rules or heuristics, which brings scalability and flexibility to NLG. Statistical models also provide an opportunity to learn in-domain human colloquialisms and cross-domain model adaptations. A robust and quasi-supervised NLG model is proposed in this thesis. The model leverages a Recurrent Neural Network (RNN)-based surface realiser and a gating mechanism applied to input semantics. The model is motivated by the Long-Short Term Memory (LSTM) network. The RNN-based surface realiser and gating mechanism use a neural network to learn end-to-end language generation decisions from input dialogue act and sentence pairs; it also integrates sentence planning and surface realisation into a single optimisation problem. The single optimisation not only bypasses the costly intermediate linguistic annotations but also generates more natural and human-like responses. Furthermore, a domain adaptation study shows that the proposed model can be readily adapted and extended to new dialogue domains via a proposed recipe. Continuing the success of end-to-end learning, the second part of the thesis speculates on building an end-to-end dialogue system by framing it as a conditional generation problem. The proposed model encapsulates a belief tracker with a minimal state representation and a generator that takes the dialogue context to produce responses. These features suggest comprehension and fast learning. The proposed model is capable of understanding requests and accomplishing tasks after training on only a few hundred human-human dialogues. A complementary Wizard-of-Oz data collection method is also introduced to facilitate the collection of human-human conversations from online workers. The results demonstrate that the proposed model can talk to human judges naturally, without any difficulty, for a sample application domain. In addition, the results also suggest that the introduction of a stochastic latent variable can help the system model intrinsic variation in communicative intention much better.

The method an individual uses to handle their passwords is an important measure to be able to protect their sensitive information from data intrusion. The amount of passwords a person uses is increasing rapidly, this leads to the individual having problems with remembering all of them. Using a password manager that stores all passwords, as well as creating unique strong passwords for different websites is a common solution for this problem.  This study aims to deepen the understanding of the individual's password management. Apple’s own password manager iCloud Keychain has been used as the platform of study. This study aims to seek answers on how secure the common user feels towards iCloud Keychain, as well as user experience point of views. How user friendly is the program and what challenges do users experience?  The study is primarily based on a questionnaire, which forms the basis for in-depth interviews with both a selected focus group as well as an expert interview.  The results show divided opinions about the level of security the users feel towards iCloud Keychain, and where connections can be seen based on users previous knowledge in the field. The conclusion includes that iCloud Keychain is quite a strong candidate on the market for password management but it has some issues when it comes to the overall user experience. Areas for development to create an even stronger candidate include amongst others, opening for use on a wider number of platforms, that is; unlocking the program from Apple's ecosystem. / Individens hantering av sina lösenord är en viktig åtgärd för att kunna skydda sin känsliga information från dataintrång. I dagsläget använder människor sig av ett större antal lösenord, vilket gör det svårare att komma ihåg dem. Att använda sig av en lösenordshanterare som samlar alla lösenord, samt även skapar unika lösenord för varje hemsida gör det säkrare för individen.  I denna studie fördjupas förståelsen för individens hantering av lösenord och dess användning av just Apples egna lösenordshanteringsprogram iCloud Nyckelring. Frågor kring användarens känsla av säkerhet gentemot programmet undersöks. Även användarvänligheten och utmaningar med användandet är några av huvudsakerna som undersöks.  Studien bygger framförallt på en enkät, som lägger grunden till en fördjupad intervju med en fokusgrupp samt en expertintervju.  Resultaten påvisar delade meningar kring säkerheten med iCloud Nyckelring, där samband kan dras utifrån om användaren har tidigare kunskap inom området eller inte. Slutsatsen inkluderar att iCloud Nyckelring är en ganska stark kandidat på marknaden för lösenordshantering, men den har vissa problem när det kommer till den övergripande användarupplevelsen. Utvecklingsområden för att skapa en ännu starkare kandidat är till exempel att öppna för användning på ett större antal plattformar, det vill säga; låsa upp programmet från Apples ekosystem.

Speech recognition is the task where a machine processes human speech into a written format. Groundbreaking scientific progress within speech recognition has been fueled by recent advancements in deep learning research, improving both key metrics of the task; accuracy and speed. Traditional speech recognition systems listen to, and analyse, the full speech utterance before making an output prediction. Streaming speech recognition on the other hand makes predictions in real- time, word by word, as speech is received. However, the improved speed of streaming speech recognition comes at a cost of reduced accuracy given the constraint of not having access to the full speech utterance at all time. In this thesis, we investigate the accuracy of streaming speech recognition systems by implementing models with state-of-the-art Transformer-based architectures. Our results show that for two similar models, one streaming, the other non-streaming, trained on a 100hr subset of Libirspeech, achieve a word error rate of 9.99%/10.76% on test- clean without using a language model. This puts the cost of streaming at a 7.2% accuracy degradation. Furthermore, the streaming models can be used “on-device” which has many benefits, including lower inference time, privacy preservation, and the ability to operate without an internet connection. / Taligenkänning är uppgiften där en dator bearbetar mänskligt tal till ett skrivet format. Forskning inom taligenkänning har drivits av de senaste framstegen inom forskning i djupinlärning, vilket har lett till att de två viktigaste mätvärdena, träffsäkerhet och hastighet, har förbättrats. Traditionella taligenkänningssystem lyssnar till och analyserar hela talsekvensen innan en prediktion görs. Strömmande taligenkänning å andra sidan gör realtids prediktioner, ord för ord, när tal tas emot. Den ökade hastigheten som strömmande taligenkänning medför kommer på bekostnad av träffsäkerhet då tillgången till hela talsekvensen inte alltid är tillgänglig. I den här avhandlingen undersöker vi träffsäkerhet av strömmande taligenkänningssystem genom att implementera ”Transformer”- baserade arkitekturer. Våra resultat visar att för två liknande modeller, en strömmande, och en icke- strömmande, tränade på 100 timmar av datasetet Librispeech, når en ordfelfrekvens på 9.99%/10.76% på ”test-clean”. Det gör att strömmande taligenkänning kommer på en bekostnad av 7.2% träffsäkerhet jämfört med icke- strömmande. De strömmande taligenkänningsmodellerna kan användas ”on-device” vilket främjar lägre slutledningstider, sekretessbevarande och förmågan att fungera utan internetanslutning.