Search

Research on autonomous obstacle avoidance of drones has recently received widespread attention from researchers. Among them, an increasing number of researchers are using machine learning to train drones. These studies typically adopt supervised learning or reinforcement learning to train the networks. Supervised learning has a disadvantage in that it takes a significant amount of time to build the datasets, because it is difficult to cover the complex and changeable drone flight environment in a single dataset. Reinforcement learning can overcome this problem by using drones to learn data in the environment. However, the current research results based on reinforcement learning are mainly focused on discrete action spaces. In this way, the movement of drones lacks precision and has somewhat unnatural flying behavior. This study aims to use the soft-actor-critic algorithm to train a drone to perform autonomous obstacle avoidance in continuous action space using only the image data. The algorithm is trained and tested in a simulation environment built by Airsim. The results show that our algorithm enables the UAV to avoid obstacles in the training environment only by inputting the depth map. Moreover, it also has a higher obstacle avoidance rate in the reconfigured environment without retraining. [ABSTRACT FROM AUTHOR]

The development of software projects requires the co-ordination of the efforts of a team of professionals with varying skills. Given the time, quality and cost constraints, the major problem in software development is to assign the right personnel to the right job at the right time and at the right cost. In this study, we propose a skill-to-time model in which the task processing time varies in accordance with the level of skill possessed by the personnel assigned to the task. We cast the problem as a cost optimization problem with several practical constraints and find an optimal solution using the Particle Swarm Optimization (PSO). [ABSTRACT FROM AUTHOR]

This paper deals with the minimization of the operational costs of business systems using the Artificial Immune System (AIS). The objective function consists of service costs and waiting costs. Service cost is the cost of hiring service-providing professionals, while waiting cost is the estimate of the loss to business as some customers, not willing to wait for service, decide to go to the competing organizations. Since business systems are inherently concurrent with multiple asynchronous activities, we have used Generalized Stochastic Petri Net as the modelling tool. The Petri net-modeled system is simulated by means of a discrete event simulator and optimized by means of the Artificial Immune System. Compared to the other meta-heuristic optimization algorithms like the Genetic Algorithm, the AIS is found to be simpler in implementation and better in performance. [ABSTRACT FROM AUTHOR]

In service systems, service cost is the cost associated with the employment of service-providing personnel, while utilization rate is the proportion of the system's resources which are used by the incoming traffic for service. Service cost and server utilization rate are conflicting objectives in service systems. In this paper, we cast the problem as a multi-objective optimization problem and use the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm to minimize the two conflicting objectives simultaneously. MOPSO is a fairly recent swarm intelligence meta-heuristic algorithm known for its simplicity in programming and its rapid convergence. The multi-objective optimization procedure is illustrated with the example of a practical service system. MOPSO produces a well-spread Pareto front for the two conflicting objective functions in the practical service system. [ABSTRACT FROM AUTHOR]

Software development projects are notorious for being completed behind schedule and over budget and for often failing to meet user requirements. A variety of cost estimation models have been proposed to predict development costs early in the lifecycle with the hope of managing the project well within time and budget. However, studies have reported rather high error rates of prediction even in the case of the well-established and widely acknowledged models. This study focuses on the improvement and fine-tuning of the COCOMO 81 model through the application of the recently developed Swarm Intelligence techniques. Recent studies have used data mining techniques to improve the prediction accuracy of COCOMO 81. Our research reconfirms these studies and makes further improvement in the prediction accuracy. The wrapper data mining method is slow and is heavily dependent on the domain experts' heuristics. The Particle Swarm Optimization (POS) meta-heuristic algorithm is fast converging and relies neither on the knowledge of the problem nor on the experts' heuristics, making its application wide and extensive. [ABSTRACT FROM AUTHOR]

The domain expert's heuristics are used in the performance design of collaborative systems which are modelled by client-server timed Petri nets. The collaborative system Petri net model is decomposed into basic primitive structures. The performance design rules that make up the expert system are categorized according to the primitive structure type. The rules are classified as: Diagnostic Rules, Primitive Structure Rules, Upstream Search Rules, Downstream Propagation Rules and Advisory Rules. In tuning the system operation, the expert system follows the policy of local improvement, while keeping the global propagation of adverse effects of tuning to a minimum. [ABSTRACT FROM AUTHOR]

Service systems are designed with the intention of providing service to customers. Also known as a value co-creation system, a service system is a configuration of technology and organizational networks. This paper describes the development of a performance ontology for service systems. The core of the performance ontology describes the four primitives in service systems – customers, resources, service and protocols, and the semantic relations among them. The ontology provides a framework for integrating the quantitative performance evaluation of service systems specified as queuing networks or as stochastic Petri nets. The goal of the domain ontology is to eliminate the conceptual and terminological confusion among the members of a virtual community of performance analysts and designers. [ABSTRACT FROM AUTHOR]

This paper proposes an analysis process and tools for analysing collaboration task. In order to analyse the collaboration task, it is important to grasp not only the workflow in the entire system, but also the flow of the personnel who are engaged in collaboration. The general workflow and the flow of personnel in collaborative systems are usually distributed and concurrent. Petri nets are ideal modelling tools for the description and analysis of concurrent systems. However, the flexibility in the Petri net constructs can at times lead to the introduction of the analyst's subjectivity in the description and analysis of a system. This fact, particularly in the case of collaborative systems, fails to depict their client-server feature. We propose disciplined client-server Peri nets and a suit of tools that aid the description, analysis and performance evaluation of collaborative systems. The Petri net Editor serves as a drawing board to sketch the system diagram; the Petri net Visualizer gives a visual display of the system operation and the Petri net Simulator simulates the system operation. By using these tools, the behaviour of the Petri net can be captured and the system performance can be estimated. We also present Petri net models of two practical collaborative systems with distributed and concurrent service and estimate their performance by using the tools that we have designed. [ABSTRACT FROM AUTHOR]

Much research has been done on Design Rationale (DR) in software development. DR consists of the decisions made during the design process and the reasons behind them. It is useful for revising, maintaining, documenting, evaluating, and learning the design. In collaboration, rational methods act as catalysts in clarifying issues and their related trade-offs for the involved participants. The three main issues in DR are the capture, representation and use of DR. However, each of these issues is rather difficult to pursue in practice mostly because of the overhead they create in the design process. In this paper, we discuss the extension of design rationale (DR) to performance design rationale (PDR) for the development of collaborative systems. The Performance Design Rationale Framework (PDRF) addresses the issues of capture, representation and use, and presents an integrated solution. In collaborative systems, near-optimum performance under the given constraints is a requirement. However, performance design is at best taken for granted, or a patch-work solution is worked out when operational problems are discovered in the deployed system. At the planning and design level, the designers hope and expect that the system will function as per the specifications after its implementation. In other words, the designer's knowledge about the system's performance is not explicit; it is only implicit. PDRF aims at capturing this tacit knowledge and representing it explicitly so that it can be used effectively. PDRF consists of three phases – (1) system modelling, (2) performance evaluation, and (3) performance improvement. It is implemented as an integrated environment with each stage seamlessly flowing into the next stage. Real-life collaborative engineering systems are too complex and too messy to allow a neat mathematical formulation. As an alternative strategy, we adopt a descriptive modelling technique known as Multi-Context Map (Map). Since collaborative systems are discrete-event in nature, performance evaluation is carried on by means of a discrete-event simulator. Finally, the performance improvement or parameter-tuning is accomplished by the application of Qualitative Reasoning (QR). The qualitative rules in the performance improvement expert system are the representation of the domain experts' heuristics or rationale for performance design. Especially through the utilization of PDRF in the development of collaborative systems, the design engineers are utilizing the performance design rationale that is embedded in the PDRF. [ABSTRACT FROM AUTHOR]

The simulated annealing (SA) algorithm has been applied for the optimization of a variety of systems. In this paper, we discuss the application of SA for optimizing the operation of collaborative systems. We choose the weighted average of the time customers spend in the system as the objective function to be minimized. This is because the time spent in the system is directly related to customer satisfaction, which in turn is related to business profits. Optimization through SA is sought for the system operating with the service time as well as the capacity of each of its servers, under constrains. We have tested the algorithm on a small-scale clinic model and found the results satisfactory. As future work, we plan to fine-tune the algorithm and apply it to larger collaborative systems. [ABSTRACT FROM AUTHOR]