MOTIVATION: Sampling k-mers is a ubiquitous task in sequence analysis algorithms. Sampling schemes such as the often-used random minimizer scheme are particularly appealing as they guarantee at least one k-mer is selected out of every w consecutive k-mers. Sampling fewer k-mers often leads to an increase in efficiency of downstream methods. Thus, developing schemes that have low density, i.e. have a small proportion of sampled k-mers, is an active area of research. After over a decade of consistent efforts in both decreasing the density of practical schemes and increasing the lower bound on the best possible density, there is still a large gap between the two. RESULTS: We prove a near-tight lower bound on the density of forward sampling schemes, a class of schemes that generalizes minimizer schemes. For small w and k, we observe that our bound is tight when k≡1(mod w). For large w and k, the bound can be approximated by 1w+k⌈w+kw⌉. Importantly, our lower bound implies that existing schemes are much closer to achieving optimal density than previously known. For example, with the current default minimap2 HiFi settings w = 19 and k = 19, we show that the best known scheme for these parameters, the double decycling-set-based minimizer of Pellow et al. is at most 3% denser than optimal, compared to the previous gap of at most 50%. Furthermore, when k≡1(mod w) and the alphabet size σ goes to ∞, we show that mod-minimizers introduced by Groot Koerkamp and Pibiri achieve optimal density matching our lower bound. AVAILABILITY AND IMPLEMENTATION: Minimizer implementations: github.com/RagnarGrootKoerkamp/minimizers ILP and analysis: github.com/treangenlab/sampling-scheme-analysis.

With AI model size doubling every four months, the size of AI compute clusters is growing rapidly to keep up. Traditional data center front-end networks rely on optical interconnect for data movement with Terabits per second of connectivity and have been doubling data rate every 4 years. As the models grow it is also expected that the back-end GPU networks expand beyond the confines of the rack. Optical interconnects will be needed at distances of tens of meters with 10Tb/s data transfer rates and doubling every 2 years will be expected. How should optical interconnects best achieve this bandwidth increase and what is the best performance axis to push next? In this panel discussion we pose the question - “Is it better to scale with more wavelengths or by adding more fibers?” Join us to see if any of the panelists can convince you they have the right approach.

With the proliferation of smart acoustic devices in everyday environments, low-power acoustic tags offer a promising choice for IoT applications. However, their highly limited operational range, throughput, and energy efficiency significantly restrict their viability for practical applications. In this work, we propose a first-of-its-kind distributed acoustic system called Disco. It consists of several low-power acoustic tags that can be flexibly composed on-demand to create an aperture array capable of distributed beamforming. The innovation of Disco lies in creating a ‘virtual’ distributed 2-speaker system that serves to wirelessly synchronize and enable distributed temporal beamforming at the tags independently. The low-power tags of Disco are prototyped with simple acoustic and analog-digital elements to create arrays comprising up to 8 tags. The beamforming performance of Disco scales with the number of tags in the array, delivering a multiple-fold increase in range, throughput, and energy efficiency. This advancement brings acoustic IoT applications closer to practical implementation.

In order to improve the size reduction of the deep learning models for deploying to edge devices, this study employs the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) method, combined with the Preserve Ratio for each layer, to formulate different compression strategies without manual parameter tuning. Additionally, it evaluates the model compression effectiveness using metrics such as Top-1 accuracy, model size, parameter count, FLOPs, and latency. The results show that compressing MobileNet-V2 at FP32 precision on the target hardware reduces inference latency by 45.8%. Moreover, the model size decreases by 79.23%, the parameter counts by 60%, and the model computational load by 90.35%.

This paper presents a method of deploying a complex machine learning model on a mobile device which has limited computing resources. With the rapid advancement of machine learning and deep learning, Natural Language Processing (NLP) has made considerable progress in recent years. The transformer approach has been widely used in NLP tasks with the ability of parallel processing in recurrent neural networks through the use of a self-attention mechanism. However, the excessive pursuit of accuracy leads to higher complexity in a model, and it becomes difficult to deploy such a model on mobile devices. To solve the problem of high computational cost, this research uses Neural Architecture Search (NAS) with Genetic Algorithm as the search strategy. It trains the supernet for performance evaluation to automatically design efficient architecture for different hardware platforms while considering network compression. We further reduce the size of the model with little impact on accuracy by applying K-means clustering. To show our method's effectiveness, we have deployed a model on a mobile device to perform machine translation. We use BLEU, FLOPs, and Latency as evaluation metrics. BLEU is used as an evaluation of sequence generative tasks while FLOPs and latency are used to observe whether the architecture found by the NAS is suitable for the hardware.

In this paper, we improve the structure of a generative adversarial network (GAN) based on three fundamental characteristics of the reflectarray for synthesizing better reflectarray patterns. For meeting various challenging requirements of beamforming, we propose a new model, which uses different tactics for improvements in the generator part of a GAN model. First, for accommodating the error-sensitive characteristics of the reflectarray antenna, a fully connected network is used to increase model precision. Second, based on the effects of interference occurring among all elements in a reflectarray, a self-attention mechanism is introduced to better handle the relationship among pixels. Third, with the needs of representing the associated values of elements based on the actual physical distance between the feed horn and elements, interpolation is used to enlarge the image, and then additional convolutional layers are provided to adjust the details. The results show acceptable radiation patterns produced by the GAN model.

Research in unmanned aerial vehicles (UAVs) has gained attention from various communities because of their potential usage in improving safety and efficiency in different applications. An UAV has shown promising results in dangerous conditions such as forest fires, search and rescue, medical deliveries, wildlife monitoring and geophysical scanning. Some external conditions like slow or no internet connection areas such as rural, farm, forest, ocean, etc. may affect the performance of the UAVs. These conditions can be considered as lost-link problems. Several approaches have been conducted to resolve such issues by implementing robust on-board architecture, machine learning approaches and developing knowledge based reasoning systems. However, much of software architecture research has concentrated on UAV implementation in normal network condition. Thus, we propose a model for considering lost-link problems in software architecture. In this paper, we describe two interconnected architectures for client and server. The UAV as a client is controlled by microkernel based architecture and the server is developed using microservice architecture. Both of them are connected using a synchronizer component to collect, filter, analyze, predict, and mitigate an UAV when a lost-link problem occurs. Therefore, the UAV can still find an appropriate action to complete a mission as far as the sensor and actuator are not in a critical condition. Experiment results show that our approach yields high percentage of mission accomplishment, fault tolerance and performance in a lost-link situation.

Involving the user in a simulated world introduces a new prospective in the agent-based simulation research. We propose a method of using a game development platform in designing a multiagent system based on the Blackboard architectural pattern, which provides a practical scenario to traffic simulation by showing the interaction between the pedestrians and the vehicles. Besides generating the log on the incidents happening and letting the user observe the simulation, our system also allows the user to get involved in the simulated world as a driver of a vehicle to interact with other cars and people walking around. The issues in analyzing the requirements of agents and implementing agents in an existing platform introduce a new challenge in software design. This work is a result of our agent-based software engineering research and reports a design method of integrating an existing platform with an architectural pattern to realize agent-based simulation in hosting different types of drivers and pedestrians in a simulated city blocks.

A smart home system can provide better services to assist users if it knows what user activities will occur beforehand. Early research in activity prediction has indicated that the result of prediction is unique, but the accuracy remains unsatisfactory if only one result is considered. To solve this problem, this paper proposes a method of leveraging multiple models. In this work, we use a Bayesian network to build a model to predict which activity will happen, and then the predicted results go through a property filtering to get the final result. Due to the possibility that residents in a smart home may have different activity patterns, we have built a Bayesian network model to learn conditional probability of resident activities from the dataset of CASAS project. At last, we compare the results and show that our method has improved coverage and accuracy in activity prediction. This proposed method belongs to an ongoing project involving learning and control in a smart home system.

This paper presents a dialog system for the user in interacting with a smart home system. Instead of using just a voice control for appliances, a dialog system gives the user a more natural way of communication with the smart home. The dialog system can take different forms of instruction from the user and provides services to the user. The user can inquire the system about some information instead of just ordering appliances to operate. With a dialog system, the system can also give a more informative response to the user. This paper focuses on processing the semantic information from the user input and producing a set of clear service request to send to the service provider. By limiting the application to a smaller set of services, we introduce an ontology reasoning technique to construct a simple dialog system that accepts explicit and implicit user requests. Two different kinds of ontology, the home ontology and the family member ontology, are constructed. A prototype system is presented along with demonstration using scenarios to show the result of our work.

The role of a service robot in a smart home can be considered as one of service providers. One difference is that this service provider can provide different services by its own and can also collaborate with others to provide different versions of services. Previous research in service provision concentrates what service devices can be integrated to introduce a new service. However, a service robot's presence can bring a new challenge to service provision. A smart home system is proposed in this paper by integrating techniques in ontology, reasoning, and planning. A user request is processed by the system to produce tasks that needs to be satisfied. A service robot performs a sequence of operations planned by hierarchical task network (HTN) planning. The use of ontology provides the definition of the context and situation in a smart home with machine-interpretable formats that will be recorded and reused. In addition, the preferences of individual users can also be recorded as personal ontology. Reasoning through case-based reasoning (CBR) determines the behaviors that a user engages in while interacting with smart homes and it also provides a learning mechanism for the interaction between the users and the smart home system.

The interaction between pedestrians and vehicles is most critical when a pedestrian tries to walk across a street. The section for a pedestrian to walk across the street is a conflict region if a vehicle is approaching that area where traffic lights are needed in order to protect pedestrians and also allow a smooth traffic flow. This research uses a multiagent method to build a simulation system to monitor how traffic lights affect the interaction between pedestrians and vehicles. By using agents, different kinds of pedestrians in different age and gender groups can be created to show a more realistic scenario of pedestrian reactions on the street. Agents control individual vehicles to simulate different driving possibilities. By using this system, we can adjust traffic lights differently to see how effective traffic lights can be set. By using agents, different research results on pedestrian models or vehicle models can be implemented independently and can be integrated in simulation. In addition, agents can also carry personal preferences to behave differently under different circumstances.

The main purpose of this paper is to provide a snapshot of the current status of our two-phase-eight-year nation-wide effort in improving the software engineering education in Taiwan. In the first phase of this program (2004 - 2008), the number of universities that regularly offer software engineering courses grew from 63 to 92 while the number of offered courses grew from 159 to 406. The main objective of the second phase (2010 - 2014) was set to establish and implement the core competences of software engineering in our module programs. Seven capabilities are identified to form the SE core competences, including think computationally, teamwork in software development and maintenance, build abstractions and perform problem domain decompositions, analyze and model complex systems involved various domains, develop, review and verify complex systems involved various domains, create user-friendly interfaces based on user experiences, and manage and evolve large-scale design and development efforts. Multiple actions have also been taken to enhance the core competences of students in ICT-related programs in Taiwan, including developing practical course material, holding training courses for the educators, providing onsite lecturing support, and delivering industrial-oriented practical courses.

A cooperative negotiation and problem solving method, ANC (Automated Negotiation and Case-based reasoning), is proposed in this paper. The goal of this method is to provide a suitable smart home service for multiple user requirements. The motivation in this paper is that agreeing on a common service is difficult when different users propose different requirements. Therefore, in ANC, cooperation negotiation is considered for resolving conflicts for making the requirements consistent among users, and based on such requirements, a common solution is provided through a reasoning process. There are five stages in ANC: issue acquisition, conflict detection, issue decision, automated negotiation, and problem solving. To make the ANC system provide personalization, a learning method based on attributes weighting has been integrated with the advantage of the constant learning ability of case-based reasoning.

We present a novel architecture to develop Virtual Environments (VEs) for multicore CPU systems. An object-centric method provides a uniform representation of VEs. The representation enables VEs to be processed in parallel using a multistage, dual-frame pipeline. Dynamic work distribution and load balancing is accomplished using a thread migration strategy with minimal overhead. This paper describes our approach, and shows it is efficient and scalable with performance experiments. Near linear speed-ups have been observed in experiments involving up to 1,000 deformable objects on a six-core i7 CPU. This approach’s practicality is demonstrated with the development of a medical simulation trainer for a craniotomy procedure.

One of attractive features of electronic-pets (e-pets) is interaction between the user and the e-pet. The interaction, however, is usually limited to using the predefined commands. In this paper, we present a way of involving the user in helping an e-pet learn high-level behaviors based on basic actions. The high-level behaviors are derived with planning, and the execution of the behaviors is then trained with reinforcement learning. In this research, we explain how we use a partially observable Markov decision process and the hierarchical task network planning for designing behaviors. A Q-learning method is then applied to the training of the e-pet for achieving the correct behavior. A prototype is presented to show its feasibility and effectiveness.

We proposed a Case-Based Planning (CBP) approach to enhance mobile agents guidance. CBP is a planning approach by reusing old plan results to solve new planning problems. There are two advantages of embed mobile agents with CBP: learning and efficiency. The agent with CBP can store successfully performed remote service requests for reuse and failed situations for avoidance. Efficiency of the agent with CBP comes from that the ability of reusing parts of old migration plans could avoid the repetition of the same efforts. It is obvious that if each agent in an agent-based environment can remember the past cooperation experience, the efficiency of agents' cooperation could be improved because more communication load and network latency can be avoided. In our approach, a plan consists of a series of actions to achieve related goals. The actions include internal services calls and external services invocations. A hierarchical case-based reasoning (HCBR) mechanism is used as core to implement our approach.

We propose a goal-driven approach to model the service request intention in service-oriented systems. The service request intention can be extracted from the user input and modeled by predefined goal models. We identify this problem as the service request intention extraction. If a service-oriented system has the abilities of user's intention extraction and can make some activities to satisfy the extracted intention, the system can provide a more convenient and efficient service for the user. We start the system construction from the view of goal-driven requirements engineering. The requirements specification is generated by the goal-based requirements analysis in which the functional and nonfunctional requirements will be extended with goal models. A set of computable goal models that represent the user requirements is selected and refined as the basis of system services. The designer can also design related system services based on the requirements specification. Based on the proposed intention extraction approach, the user's vague and imprecise intention will be extracted and mapped to computer understandable and computable goal models for representing the intention. A case-based method is developed to implement the intention extraction process. The intention interpretation knowledge is stored in a case base, and the intention interpretation is based on the process of case retrieval and adaptation. A general architecture for an intention-aware service-oriented system is proposed for demonstrating how to apply the proposed approach. © 2010 Wiley Periodicals, Inc.

We propose a goal-driven approach to model explicit and implicit service requests for intelligent interface design. This approach starts the system construction from the view of software requirements engineering. Requirements of the system are generated by the goal-based requirements analysis in which functional and nonfunctional requirements will be represented by a set of goal models. These goal models are used to represent possible explicit and implicit requests in the userpsilas service requests. The goal models will be future selected and refined by performing a proposed service request interpretation process. The process is used to extract the userpsilas explicit and implicit service request in an input service request phrase. A service request interpreter is implemented to demonstrate the proposed approach.

We propose a case-based service request interpretation approach for digital homes. This approach starts the system construction from the view of software requirements engineering and a goal-driven approach is used to model service requests. The predefined goal model will be encapsulated in a case of a case base. The cases will be future selected and refined by performing a proposed case-based service request interpretation process. The system will base on the selected case to perform a series of actions to satisfy the service requests. Based on the case adaptation process, the system can adapt different user's service request and give a convenient interface for digital homes. We give some scenarios to demonstrate applications of the proposed approach.

Considering user preferences in defining robot services is important. How to interact with service robots varies from user to user, and providing a user friendly way for the users to interact with robots becomes necessary since the users are usually non-technical people when dealing with service robots. This article focuses on defining robot services that can suit user's preferences. Service robots are typically designed to provide one set of services not targeted to a particular user. However, ways of controlling robots may differ depending on users. Learning from past experiences enables a robot to adapt to the user's specific needs and interacting style. In this research, we use ontology as the design tool for defining robot services and uses case-based reasoning as a means of learning in storing previous interaction experiences as cases. These cases are reused the next time when similar requests are made.

We propose an integer programming (IP) framework to undertake the dedicated photolithography machine constraint in semiconductor manufacturing. The constraint is one of the new challenges set by the process engineer in semiconductor manufacturing due to natural bias of photolithography machines. Previous researches either did not take the constraint into account or the proposed heuristic approach might not efficiently fit the fast-changing market of semiconductor manufacturing. In this paper, the proposed IP framework provides an approach to minimize the production cost in an efficient time. We also present the experiments to validate the approach.

We proposed a Case-based planning (CBP) approach for agent-based service-oriented systems. CBP is planning by reusing old plan results. It reuse past successfully executed plans to solve new planning problems. There are two advantages of CBP: learning and efficiency. CBP can remember successfully performed cases for reuse and failed situations for avoidance. It can also remember repairs strategies for reapplication. Efficiency of CBP comes from that the ability of reusing parts of old plans could avoid the repetition of the same efforts. It is obvious that if each agent in the agent-based service-oriented system can remember the past cooperation experience, the efficiency of agents' cooperation could be improved because a lot of communication load and network latency can be avoided. In our approach, a plan consists of a series of actions to achieve related goals. The actions include internal function calls and external services invocations. A hierarchical case-based reasoning (HCBR) mechanism is used as core to implement CBP. We discussed how the proposed approach be used to enhance the service assessment.

For the lane-change maneuvers under the architectures of automated highway systems, most researchers focus on the maneuvers for a single vehicle to change lanes. However, not only are the lane-change maneuvers for a single vehicle needed, but a platoon-lane-change (PLC) maneuver is also required in some situations for an entire platoon to change lanes. The goal of this paper is to design PLC maneuvers under the coordinated- and noncoordinated-platooning infrastructures. Two PLC maneuvers are proposed in this paper, and they are the leader and predecessor PLC maneuvers, where the predecessor PLC maneuver is considered in two cases, i.e., the intraplatoon spacing less than or greater than the minimum safety spacing. We use SmartAHS to simulate these PLC maneuvers, and the simulation results demonstrate the feasibility of these PLC maneuvers while taking the safety and comfort of passengers into account.

This paper describes real-time volumetric haptic and visual algorithms developed to simulate burrhole creation for a virtual reality-based craniotomy surgical simulator. A modified Voxmap point-shell algorithm (McNeely et al., 1999), (Renz et al., 2001) is created to simulate haptic interactions between bone cutting tools and voxel-based bone. New surface boundary detection and force feedback calculation methods help reduce "force discontinuities" of the original Voxmap point-shell algorithm. To maintain stable haptic update rates, new forces are calculated outside the haptics rendering loop. A multi-rate haptic solution (Cavusoglu and Tendick, 2000) is used to introduce calculated forces into the haptics loop and to interpolate forces between updates. A bone erosion method is also created to simulate bone drilling capabilities of different tools. 3D texture-based volume rendering is used to display the bone and to visually remove bone material due to drilling in real-time. Volumetric shading is computed by the GPU of the video card. The algorithms described make it possible to simulate several tools typically used for a craniotomy. Realistic 3D models are also created from real surgical tools and controlled by the haptic device

In order to meet new mission requirements or to adapt environmental changes, robots perform a task switch or role switch. When a robot is required to perform a specific task, which is beyond its skills, it might need to request other robots' help. In order to improve the flexibility, we use the concept of a software mobile agent to design an architecture called Virtual Operator MultiAgent System (VOMAS). In VOMAS, intelligent agents, called a virtual operator (VO) and a robot agent (RA), work together to control a robot to fulfill a specific task. Each task is represented by a VO and the RA handles reactive control. Based on this architecture, VOMAS can perform the dynamic task switch to handle missions, which are not initially given. Two case studies, a wheelchair and formation control, are introduced for explaining our approach and to show the feasibility and effectiveness of dynamic task switch. Furthermore, VOMAS can be treated as a telerobotic architecture, and the network-load test shows that VOMAS requires less network load than direct control

We propose a method which reuse an old plan from past experiences to satisfy the newly created service request from the user. The user's service request intention is interpreted and represented by computable goal models. A plan consists of a series of actions to achieve related goal models. The actions include internal function calls and external services invocations. A hierarchical case-based reasoning (HCBR) mechanism is used to store and reuse the past problem solving experiences. When new goal models are identified, the HCBR retrieves an old plan and adapt the plan for achieving the newly identified goal. The case base is decomposed into two levels: abstract level and concrete level. The case in the abstract level contains the descriptions of sub-problems, and the case in the concrete level contains system functions to solve the sub-problems. The proposed approach is implemented in the service requester agent of the Advanced Traveler Information Systems (ATIS) to enhance the service assessment.

We present a multiagent scheduling (MS) system to tackle the dedicated machine constraint in this paper. The dedicated machine constraint is one of the new issues of the photolithography machinery due to natural bias. Natural bias will impact the alignment of patterns between different photolithography layers. The dedicated machine constraint is the most important challenge to improve productivity and fulfill the request for customers in semiconductor manufacturing today. In this paper, the proposed MS system is based on a resource schedule and execution matrix (RSEM) and keeps the load balancing among photolithography machines during each scheduling step according to the current load among the photolithography machines in the production line. We describe the prototype system including the agents and the coordination strategies in the paper. We also demonstrate the simulation result that validated the proposed MS system.

We propose a heuristic load balancing (LB) scheduling approach based on a resource schedule and execution matrix (RSEM) to tackle the dedicated machine constraint for the photolithography process in semiconductor manufacturing. The constraint of having a dedicated machine is one of the new challenges introduced in photolithography machinery due to natural bias. With this dedicated machine constraint, if we randomly schedule the wafer lots to arbitrary photolithography machines at the first photolithography stage, then the load of all photolithography machines might become unbalanced. However, many scheduling policies or modeling methods proposed by previous research for the semiconductor manufacturing production have not discussed this dedicated machine constraint. In this paper, along with providing the LB approach to the issue of the dedicated machine constraint, we also present a novel model-the representation and manipulation methods for the task patterns. The advantage of LB is to easily schedule the wafer lots by simple calculation on a two-dimensional matrix. We present the result of the simulations to validate our approach as well.

A fuzzy time series data representation method based on the Japanese candlestick theory is proposed and used in assisting financial prediction. The Japanese candlestick theory is an empirical model of investment decision. The theory assumes that the candlestick patterns reflect the psychology of the market, and the investors can make their investment decision based on the identified candlestick patterns. We model the imprecise and vague candlestick patterns with fuzzy linguistic variables and transfer the financial time series data to fuzzy candlestick patterns for pattern recognition. A fuzzy candlestick pattern can bridge the gap between the investors and the system designer because it is visual, computable, and modifiable. The investors are not only able to understand the prediction process, but also to improve the efficiency of prediction results. The proposed approach is applied to financial time series forecasting problem for demonstration. By the prototype system which has been established, the investment expertise can be stored in the knowledge base, and the fuzzy candlestick pattern can also be identified automatically from a large amount of the financial trading data.

A goal driven intention extraction approach is proposed to automate the process of extracting user intention from the original Web Service query terms. The input query terms will be parsed into different word sense sets. By the domain ontology, possible senses of the terms will be identified. A goal structure is constructed to help identification of goal models. Combining the information of query terms and goal structures, one or more goal models will be identified. A goal selector will select a candidate goal model from generated goal models to represent the user intention. A BDI agent is used as framework to implement the intention extraction tool.

A method to integrate the Web services, agent skills, and human information for the user's intentions satisfaction is proposed and discussed. The user's intention is extracted and represented by goal models. This makes the imprecise intentions be evaluated. Based on the hierarchical task network (HTN) planning approach, a plan could be automatically generated for satisfying the user's intention extracted from the entered service request string. A bookstore scenario is used to illustrate the proposed approach and a service requestor agent is introduced for implementation.

An intelligent tool for the investment knowledge learning and sharing is proposed and developed. The candlestick pattern is a widely used empirical model of investment decision and the investors can make investment decisions by the identified candlestick patterns. The imprecise and vague investment knowledge is represented in fuzzy candlestick patterns. The pattern consists of candlestick lines which are derived from the variation of the trading prices. The investor can create, edit, validate, and share the patterns by the proposed system. A graphical user interface (GUI) is designed to display the candlestick patterns and the user can also capture the pattern directly from the GUI. A fuzzy pattern recognition module is implemented in the system to help the user to identify the patterns automatically from the large amount financial database.

Designing cooperative multi-robot systems (MRS) requires expert knowledge both in control and artificial intelligence. Formation control is an important research within the research field of MRS. Since many researchers use different ways in approaching formation control, we try to give a taxonomy in order to help researchers design formation systems in a systematical way. We can analyze formation structures in two categories: control abstraction and robot distinguishability. The control abstraction can be divided into three layers: formation shape, reference type, and robotic control. Furthermore, robots can be classified as anonymous robots or identification robots depending on whether robots are distinguishable according to their inner states. We use this taxonomy to analyze some ground-based formation systems and to state current challenges of formation control. Such information becomes the design know-how in developing a formation system, and a case study of designing a multi-team formation system is introduced to demonstrate the usefulness of the taxonomy

The concept of knowledge-based software architecture has recently emerged as a new way to improve our ability to effectively construct and maintain complex, large-scale software systems. Under this new paradigm, software engineers are able to do evolutionary design of complex systems through architecture specification, design rationale capture, architecture validation and verification, and architecture transformation. This paper surveys some of the important techniques that have been developed to support these activities. In particular, we are interested in knowledge/requirements acquisition and analysis. We survey some tools that use the knowledge-based approach to solve these problems. We also discuss various software architecture styles, architecture description languages (ADLs) and features of ADLs that help build better software systems. We then compare various ADLs based on these features. The efficient methods that were developed for verification, validation and high assurance of architectures are also discussed. Based on our survey results, we give a basis for comparing the various knowledge-based systems and list these comparisons in the form of a table.

For mobile robot navigation, controlling the components such as sensors and motors is not enough. Besides controlling such components, a navigation control system needs to provide decisions in choosing a path to go and carrying out requests from the user. There are many tasks which need to be performed concurrently. Intelligent agents are suitable for being responsible for carrying out each task and performing cooperation between the agents. The paper describes an architecture which uses agents in a cooperative environment.

In software engineering, requirements analysis is a knowledge intensive task: and it requires an expert to understand what the clients need. We introduce a method which contains different AI techniques to perform this task, and a knowledge-based requirements analysis system, RAKES is presented to explain our approach. With RAKES, not only the ordinary functional requirements are collected, but also the non-traditional information. Nonfunctional requirements include the quality of operations or the background information for constructing the requirements which are gathered through a knowledge-based support. Different kinds of information collected are stored and organized in a knowledge base and can be used as the source of the user input in the latter phases of software development. Algorithms and procedures have been developed for constructing the interface language, organizing the knowledge base, and applying the knowledge base to different tasks. RAKES is integrated to the FRORL architecture to offer a systematic way toward requirements analysis, specification production, prototype generation, specification debugging, and code transformation.

Current requirements specifications tend to focus more on the functional aspect than the nonfunctional side of a product. Many techniques and tools have been developed to specify and evaluate functional requirements, but few of them are equipped to deal with nonfunctional requirements. In this paper we extend the formal requirements specification language, FRORL, to model nonfunctional requirements and show how these nonfunctional requirements are related to the functional requirements. We also introduce a parallel evaluation technique to evaluate the functional requirements model by satisfying the nonfunctional requirements associated to it. By examining the result from the functional model, we can see how the nonfunctional requirements are satisfied, so that we can adjust and modify the nonfunctional requirements.>

Safety-critical systems are not only difficult to build, but also difficult to debug because they often have strict timing constraints and non-deterministic behavior. A correct and precise specification reduces the effort spent in testing and debugging the implemented system. This paper presents techniques of specification debugging and issues related to it. We introduce an approach to the debugging of a specification in FRORL (Frame and Rule Oriented Requirements Language), which supports non-determinism and non-monotonicity in a system. The approach aids the user in detecting and correcting the possible faults which can arise not only when writing the specification, but also after the verification of the specification.

The use of ray casting in volume rendering and its uses and advantages over surface rendering algorithms are discussed. Various adaptive algorithms that attempt to overcome its problem of high computational cost by taking advantage of image coherency and the bandlimited nature of volume data are described. A method of subdividing the image plane with isosceles triangles, instead of quadrants as is usually done is proposed. It results in fewer rays being fired without sacrificing image quality. A brief theoretical analysis of the algorithm in comparison with other methods is given.>