Constructing and maintaining a formation is critical in applications of cooperative control of multi-agent systems. In this research we address the formation control problem of generating a formation for a group of nonholonomic mobile agents. The formation control scheme proposed in this work is based on a fusion of leader-follower and virtual reference approaches. This scheme gives a formation constraint representation that is independent of the number of agents in the formation and the resulting control algorithm is scalable. One of the important desired features in controller design is that the formation errors defined by formation constraints should be stabilized globally and exponentially by the controller. The proposed controller is based on feedback linearization, and formation errors are shown to be globally exponentially stable in the sense of Lyapunov. Since formation errors are stabilized globally, the proposed controller is applicable to both formation keeping and formation construction problems. As a possible application, the proposed algorithm is implemented in a cooperative ground moving target surveillance scenario. The proposed algorithm enables the determination of the minimal number of agents required for surveillance of a moving target. The number of agents returned by this scheme is not optimal and hence is a conservative solution. However, this is justified by the computational savings the scheme offers.

Reinforcement learning is a machine learning technique designed to mimic the way animals learn by receiving rewards and punishment. It is designed to train intelligent agents when very little is known about the agent’s environment, and consequently the agent’s designer is unable to hand-craft an appropriate policy. Using reinforcement learning, the agent’s designer can merely give reward to the agent when it does something right, and the algorithm will craft an appropriate policy automatically. In many situations it is desirable to use this technique to train systems of agents (for example, to train robots to play RoboCup soccer in a coordinated fashion). Unfortunately, several significant computational issues occur when using this technique to train systems of agents. This dissertation introduces a suite of techniques that overcome many of these difficulties in various common situations. First, we show how multi-agent reinforcement learning can be made more tractable by forming coalitions out of the agents, and training each coalition separately. Coalitions are formed by using information-theoretic techniques, and we find that by using a coalition-based approach, the computational complexity of reinforcement-learning can be made linear in the total system agent count. Next we look at ways to integrate domain knowledge into the reinforcement learning process, and how this can signifi-cantly improve the policy quality in multi-agent situations. Specifically, we find that integrating domain knowledge into a reinforcement learning process can overcome training data deficiencies and allow the learner to converge to acceptable solutions when lack of training data would have prevented such convergence without domain knowledge. We then show how to train policies over continuous action spaces, which can reduce problem complexity for domains that require continuous action spaces (analog controllers) by eliminating the need to finely discretize the action space. Finally, we look at ways to perform reinforcement learning on modern GPUs and show how by doing this we can tackle significantly larger problems. We find that by offloading some of the RL computation to the GPU, we can achieve almost a 4.5 speedup factor in the total training process.

This thesis presents the research results of a correlation and model based on nano and macroindentation hardness measurements. The materials used to develop and test the correlation include bulk tantalum and O1 tool steel. Following the literature review and a detailed description of the experimental techniques, the results of the nanoindentation hardness measurements are presented. After applying the methods and correlation recommended here, the results should give an accurate value of hardness in the Vickers scale for microstructural features that are too small to be precisely and exclusively measured using the traditional macroindentation hardness technique. The phenomena and influential factors in nanoindentation hardness testing are also discussed. These phenomena and theories are consistent with the microstructural behavior predicted in the Nix and Gao model for mechanism-based strain gradients. Implementing the correlation factors and/or correlation curve, accurate results can be found for metals over a broad hardness range. Initially, this research may impact the pipeline division of the petroleum industry by providing a correlation to the Vickers scale for nanoindentation testing of microstructural features. This thesis may also provide a research methodology to develop hardness correlations for materials other than metals. This thesis consists of eight chapters. Following an introduction in Chapter I, the research motivations and objectives are highlighted in Chapter II. Chapter III explains the multi-scale indentation techniques used in this thesis and Chapter IV presents the materials preparation techniques used. Then, the results are presented in Chapter V, followed by the factors affecting nanoindentation hardness in Chapter VI. Finally, Chapters VII and VIII reveal the indentation contact analysis, correlation, and conclusions of this research, respectively.

A multi-speed drive hub is one of important components in transmission system. The purpose of this work is to develop a system methodology for the design of multi-speed bicycle drive hubs. First, analyse multi-speed drive hub patents to identify the basic characteristics and design requirement of multi-speed drive hub. Second, using a systematized mathod to synthesize the feasible gear trains. Third, a method is proposed to determine the feasible multi-speed sequence tables for each gear train. Fourth, an analytic method is formulated to synthesize the gear ratios and gear teeth according to a set of desired speed ratios. Fifth, design and dispose governors of multi-speed drive hubs. The results of this work will be benefit to the development and design of multi-speed bicycle drive hubs.

To support the transmission of real-time multimedia stream, Stream Control Transmission Protocol (SCTP) developed by IETF is considered to be more efficient because of its high-degree expandability and compatibility. Today we can observe that instead of using SCTP may become the transmission protocol of next-generation IP network. In this Thesis, we propose a mechanism to upgrade TCP and UDP, the multi-home feature of SCTP to ensure that multiple paths can exist between two SCTP ends. Not only can the primary path continue to function, but the secondary paths covey part of data packets once the network congestion occurs. Considering the dynamic change of our Internet, the proposed mechanism can enhance the effectiveness of SCTP data transmission, and increase overall network utilization. Cutting user data into chunks in SCTP, we can analyze the transmission performance of individual path by measuring the transmission delay from the sender to the receiving end. By modifying the simulator of NS-2, we set up different topologies in the experiment to analyze the performance of our mechanism. We compare the modified SCTP with the original SCTP to highlight our proposed mechanism in increasing throughput and network utilization by adjusting the background traffic on the paths.

Abstract Small and medium enterprises (SMEs) were one of the major contributors to the economic miracle of Taiwan. There are over a million SMEs, which comprise 97.7% of all Taiwanese enterprises, However, SMEs faced with high risks of failure, especially at their entrepreneurship stages, due to their lack of expertise and abilities in technologies developments, together with their disadvantages in economy scale, market, capital and business connections. Thus more than 60 incubators were founded by Taiwan Government in past seven years to help SMEs entrepreneur cope with the above threaten. These incubators vary in factors such as capabilities, facility, and resources. Therefore, the selection of an appropriate incubator is a critical issue for an SME entrepreneur. This research identifies a set of evaluating criteria within five categories: facility, technical and human resource, business administrative, knowledge source, and organization infrastructure based on extensive literature review and interview with industry experts. ELECTRE I, a multiple criteria decision making method is applied to evaluate incubators. The results provide an efficient and effective method for selecting an appropriate incubator for SME entrepreneur.

This paper presents design and fabrication of ring cavity semiconductor lasers and optical filters with simple fabrication process. A 1.55

The spectral range of quaternary AlGaInN/AlGaN MQWs extends from UV to IR. Nitride-based green and blue LEDs reveal a high efficiency for the further application. Integrating LEDs of three element colors can perform white light. The optical properties of GaN MQWs are very sensitive to the growth conditions of MQWs. The ununiformity is not fabrication desired but needs to prevent, which is necessary to understand and to precisely control through its growth condition for manufacture the LED. In this work the sample has a luminescence varied from orange to purple across the whole wafer. In this work, the correlations between optical and structural properties in these samples have been studied by means of Transmission Electron Microscopy (TEM), energy dispersive X-ray spectrometry (EDS), and cathodoluminescence (CL) measurements.

Bluetooth is a low-cost, low-power and short-range wireless communication technology, which operates in the global unlicensed 2.4GHz ISM band. Recently many people have paid close attention to do a lot of relevant research and product development on it. In addition to the scatternet formation problem, two major issues of Bluetooth, scheduling and routing, also greatly interest the researchers. In this paper, we propose new scatternet topology architecture for Bluetooth networks, called multi-ring scatternet. The scatternet formation and packet routing method are presented. The self-routing property of multi-ring scatternet makes its routing method efficient and low overhead. Comparing to the Bluetree scatternet, the proposed scatternet is more reliable and relieves the bottleneck at root nodes. As compared with the BlueRing scatternet, it reduces the network diameter and average routing path. The simulation results reveal that our multi-ring scatternet topology gains better performance by the three metrics of number of piconets, average routing path length and throughput.

The laser-induced holographic grating technique was employed to study the dynamic of the intensity grating formation in the azo-dye doped liquid crystals. The liquid crystal material in this study has several mesomorphic phases: Smectic C, Smectic A, Nematic and Isotropic. The first order of diffraction in the mesomorphic phases have been investigated by changing the polarizations of the probe beam.