教学课程资源库 更多>>

关键词： Computer science   Robotics   Artificial intelligence  

摘要： Enabling robots to physically engage with their environment in a safe and efficient manner is an essential step towards human-robot interaction. To date, robots usually operate as pre-programmed workers that blindly execute tasks in highly structured environments crafted by skilled engineers. Changing the robots’ behavior to cover new duties or handle variability is an expensive, complex, and time-consuming process. However, with the advent of more complex sensors and algorithms, overcoming these limitations becomes within reach. This work proposes innovations in artificial intelligence, language understanding, and multimodal integration to enable next-generation grasping and manipulation capabilities in autonomous robots. The underlying thesis is that multimodal observations and instructions can drastically expand the responsiveness and dexterity of robot manipulators. Natural language, in particular, can be used to enable intuitive, bidirectional communication between a human user and the machine. To this end, this work presents a system that learns context-aware robot control policies from multimodal human demonstrations. Among the main contributions presented are techniques for (a) collecting demonstrations in an efficient and intuitive fashion, (b) methods for leveraging physical contact with the environment and objects, (c) the incorporation of natural language to understand context, and (d) the generation of robust robot control policies. The presented approach and systems are evaluated in multiple grasping and manipulation settings ranging from dexterous manipulation to pick-and-place, as well as contact-rich bimanual insertion tasks. Moreover, the usability of these innovations, especially when utilizing human task demonstrations and communication interfaces, is evaluated in several human-subject studies.

关键词： Physics   Computational physics   Computer science  

摘要： In this thesis, we study a broad range of physical phenomena from the perspectives of theory-driven, and machine learning models. We begin by introducing a generalization of the Momentum Average method for finding numerically exact Green's functions of arbitrary polaron systems at zero and finite temperature. This method utilizes the physical ansatz that phonons are produced largely in clouds, and systematically constructs a closure of auxiliary Green's functions to ultimately solve for the spectrum. We seamlessly apply this method to a variety of problems, including the Holstein, Peierls, and mixed-boson mode models. Next, we leverage fundamental quantum mechanics to develop a microscopic model of exciton and trion scattering in monolayer transition metal dichalcogenides. We conclude that elastic scattering mechanisms are largely the dominant contributor, and confirm that our calculated doping-dependent linewidths qualitatively agree with experiment. In addition, we use Monte Carlo dynamics to examine entropically activated dynamics in continuous phase space models, and show that global and local dynamics both exhibit entropy driven activation. The second type of work discussed in this thesis pertains to data-driven machine learning models. These approaches offer the utility of instantaneous inference, which has tremendous potential application in applied science in areas such as surrogate modeling and creating digital twins of expensive experiments. First, we demonstrate that x-ray absorption spectra can be used to classify absorbing sites' local atomic information, specifically its coordination number. Next, we show that graph-based neural networks can to quantitative accuracy, predict the x-ray absorption spectrum of small molecules in the QM9 database. We highlight the various ways in which these types of methodologies can be applied to e.g. closing the design loop and surrogate modeling in general.

关键词： 多样性   理论计算机科学   算法   体外化   人类世  

关键词： Artificial intelligence   Robotics   Computer science   Violations   Algorithms   Sensitivity analysis   Temporal logic   Experiments   Case studies  

摘要： Autonomous Vehicles (AV) are inevitable entities in future mobility systems that demand safety and adaptability as two critical factors in replacing/assisting human drivers. Safety arises in defining, standardizing, quantifying, and monitoring requirements for all autonomous components. Adaptability, on the other hand, involves efficient handling of uncertainty and inconsistencies in models and data. First, I address safety by presenting a search-based test-case generation framework that can be used in training and testing deep-learning components of AV. Next, to address adaptability, I propose a framework based on multi-valued linear temporal logic syntax and semantics that allows autonomous agents to perform model-checking on systems with uncertainties. The search-based test-case generation framework provides safety assurance guarantees through formalizing and monitoring Responsibility Sensitive Safety(RSS) rules. I use the RSS rules in signal temporal logic as qualification specifications for monitoring and screening the quality of generated test-drive scenarios. Furthermore, to extend the existing temporal-based formal languages’ expressivity, I propose a new spatio-temporal perception logic that enables formalizing qualification specifications for perception systems. All-in-one, my test-generation framework can be used for reasoning about the quality of perception, prediction, and decision-making components in AV. Finally, my efforts resulted in publicly available software. One is an offline monitoring algorithm based on the proposed logic to reason about the quality of perception systems. The other is an optimal planner (model checker) that accepts mission specifications and model descriptions in the form of multi-valued logic and multi-valued sets, respectively. My monitoring framework is distributed with the publicly available S-TaLiRo and Sim-ATAV tools.

关键词： Computer science   Information technology   Artificial intelligence  

关键词： Computer science   Computer engineering   Artificial intelligence  

摘要： Arrhythmia is the irregular rhythm of a problematic heartbeat that is caused by an issue with the electrical impulses that control the heart. These issues can lead to harmful cardiovascular deceases if not diagnosed and treated. In order to detect such irregularities in the heart, the resulting electric impulses from the contraction and relaxation of the heart muscles can be detected with the use of an Electrocardiograph which is examined by a trained professional. These heartbeats follows a general pattern with a priority in focus around the QRS complex. This pattern can be leveraged to train a neural network for anomaly detection by detecting when the impulses are different from what is expected for a normal heartbeat pattern. To extract these patterns, small sections of the time-series electrocardiographs will be extracted around the QRS complex and prepared by de-noising the signal with a bandpass filter, then normalized using an L2 normalization algorithm before being passed to a neural network. This neural network utilizes a Long-Short-Term-Memory layer as its core and is capable of classifying the differences in the heartbeat as one of eighteen different heartbeats with an accuracy of 97\% within 55ms when ran on a single-board computer like a Raspberry Pi 4. This demonstrates that a neural network can be used along side a portable electrocardiogram to provide real time diagnosis on arrhythmia without the requirement of lengthy hospital stays or expensive equipment as the neural network can keep up with the heart-rate of an individual while running on cheap, portable computing systems.

关键词： Computer engineering   Computer science   Artificial intelligence   Propagation   Neurons   Accuracy   Embedded systems   Deep learning   Neural networks   Dissertations & theses   Design   Noise   Electrical engineering   Field programmable gate arrays  

关键词： Energy   Computer science   Materials science   Mechanics   Optics   Physical geography   Physics  

摘要： The aim of this thesis is to advance a fundamental understanding of process mechanics and surface integrity in micro-grinding of advanced materials, such as additively and conventionally manufactured titanium and engineering ceramic. Grinding forces and specific grinding energy were chosen as the two most important indicators to investigate the process. The surface integrity was evaluated using X-ray diffraction measurements to quantify residual stresses, surface roughness measurements, SEM microscopy, and confocal images. At the first stage, the influence of different micro-grinding and dressing parameters and different tool specifications was examined. Then, via process modelling, the outcome of the micro-grinding process at different chip thicknesses and aggressiveness numbers was studied. Additionally, a hybrid laser-assisted micro-grinding process was developed to improve the process efficiency. The results show that the flow stress of the material did not change with the changing of cutting speed and cutting strain rate. Moreover, it was observed that the depth of cut and grinding feed rate had the same (neutral) effect on the resultant grinding forces. Therefore, the efficiency of titanium micro-grinding could be highly influenced by changing the topography of the micro-grinding tool through different dressing parameters. However, using higher chip thicknesses resulted in a more efficient process in terms of cutting/chip-formation. The lowest specific energy, obtained in the single grain tests, was 11.5 J/mm3 for both-types of titanium materials. In contrast, a much higher minimum specific energy in real micro-grinding process with several (bonded) grains was observed – showing a higher amount of ploughing and rubbing/friction in the micro-grinding process. The build-up direction of additively-manufactured titanium, at low chip thickness, affected the process efficiency. In larger chip thicknesses, almost the same specific energy was measured independent of t

关键词： Computer science   Mathematics   Computer engineering  

摘要： The Rivest-Shamir-Adleman (RSA) is a public key cryptosystem widely used in online transactions over the Internet to secure the data transmission. The security of RSA relies on the computational complexity of the large integer factorization problem, for which no efficient algorithm is known. This paper proposes an easy and efficient n factoring algorithm for RSA breaking. The first step is to reduce the loop times of the algorithm (such as from O(n) to O(√n)). The preliminary study is that p×q=n can be transformed into a Diophantine equation. And it has been performed by using different integer factorization algorithms in C language. The second step is to speed up the computing in each iteration. To support the second step, Field Programmable Gate Array (FPGA) it may need to use to implement a customized multiple-core processor, which can do parallel computing for the calculation. The focus of this paper is on the first step, reducing the loop times of the algorithm, and that process has been presented in a more visualized way. Keywords: RSA, Integer factorization, Large number factoring, Cyber Security, Primality test, Fermat’s Algorithm, C language.

关键词： Cryptocurrency   Blockchain   Smart contracts   Bitcoin   Security   Peer-to-peer computing   Computer science   Blockchain   cryptocurrency   exchange   Ethereum   smart contracts  

摘要： Inspired by Bitcoin, many different kinds of cryptocurrencies based on blockchain technology have turned up on the market. Due to the special structure of the blockchain, it has been deemed impossible to directly trade between traditional currencies and cryptocurrencies or between different types of cryptocurrencies. Generally, trading between different currencies is conducted through a centralized third-party platform. However, it has the problem of a single point of failure, which is vulnerable to attacks and thus affects the security of the transactions. In this paper, we propose a distributed cryptocurrency trading scheme to solve the problem of centralized exchanges, which can achieve secure trading between different types of cryptocurrencies. Our scheme is implemented with smart contracts on an Ethereum blockchain and deployed on an Ethereum test network. In addition to implementing transactions between individual users, our scheme also allows transactions among multiple users. The experimental result proves that the cost of our scheme is acceptable.