基于改进D?Lite遗传算法路径规划研究

针对遗传算法在移动机器人路径规划中易产生早熟现象和收敛速度慢的问题,提出了改进的D~* Lite遗传算法。该算法将D~* Lite算法和遗传算法相结合,通过引入碰撞系数和可视检测技术以提高路径安全性,寻找最短路径。在遗传算法设计中加入动态调整交叉与变异概率,以解决算法在路径规划中因陷入局部最优值而不能到达目标点的问题。最后,通过实验仿真可知:与蚁群算法和免疫遗传算法相比,改进的D~* Lite遗传算法执行效率高,可以快速规划出全局最优路径。     

A lightweight autonomous MAV for indoor search and rescue

Micro Aerial Vehicles (MAVs) have great potentials to be applied for indoor search and rescue missions. In this paper, we propose a modular lightweight design of an autonomous MAV with integrated hardware and software. The MAV is equipped with the 2D laser scanner, camera, mission computer and flight controller, running all the computation onboard in real time. The onboard perception system includes a laser‐based SLAM module and a custom‐designed visual detection module. A dual Kalman filter design provides robust state estimation by multiple sensor fusion. Specifically, the fusion module provides robust altitude measurement in the circumstance of surface changing. In addition, indoor‐outdoor transition is explicitly handled by the fusion module. In order to efficiently navigate through obstacles and adapt to multiple tasks, a task tree‐based mission planning method is seamlessly integrated with path planning and control modules. The MAV is capable of searching and rescuing victims from unknown indoor environments effectively. It was validated by our award‐winning performance at the 2017 International Micro Air Vehicle Competition (IMAV 2017), held in Toulouse, France. The performance video is available on https://youtu.be/8H19ppS_VXM .     

Revisiting the Parallel Strategy for DOACROSS Loops

Journal of Computer Science and Technology - DOACROSS loops are significant parts in many important scientific and engineering applications, which are generally exploited pipeline/wave-front...     

A web-based software tool for participatory optimization of conservation practices in watersheds

WRESTORE (Watershed Restoration Using Spatio-Temporal Optimization of Resources) is a web-based, participatory planning tool that can be used to engage with watershed stakeholder communities, and involve them in using science-based, human-guided, interactive simulation–optimization methods for designing potential conservation practices on their landscape. The underlying optimization algorithms, process simulation models, and interfaces allow users to not only spatially optimize the locations and types of new conservation practices based on quantifiable goals estimated by the dynamic simulation models, but also to include their personal subjective and/or unquantifiable criteria in the location and design of these practices. In this paper, we describe the software, interfaces, and architecture of WRESTORE, provide scenarios for implementing the WRESTORE tool in a watershed community's planning process, and discuss considerations for future developments.     

Managing readiness-relevant activities for the organizational dimension of ERP implementation

An organization requires performing readiness-relevant activities to ensure successful implementation of an enterprise resource planning (ERP) system. This paper develops a novel approach to managing these interrelated activities to get ready for implementing an ERP system. The approach enables an organization to evaluate its ERP implementation readiness by assessing the degree to which it can achieve the interrelated readiness relevant activities using fuzzy cognitive maps. Based on the interrelationship degrees among the activities, the approach clusters the activities into manageable groups and prioritizes them. To help work out a readiness improvement plan, scenario analysis is conducted.     

Electromagnetism-like algorithms for optimized tool path planning in 5-axis flank machining

Optimization of tool path planning using metaheuristic algorithms such as ant colony systems (ACS) and particle swarm optimization (PSO) provides a feasible approach to reduce geometrical machining errors in 5-axis flank machining of ruled surfaces. The optimal solutions of these algorithms exhibit an unsatisfactory quality in a high-dimensional search space. In this study, various algorithms derived from the electromagnetism-like mechanism (EM) were applied. The test results of representative surfaces showed that all EM-based methods yield more effective optimal solutions than does PSO, despite a longer search time. A new EM-MSS (electromagnetism-like mechanism with move solution screening) algorithm produces the most favorable results by ensuring the continuous improvement of new searches. Incorporating an SPSA (simultaneous perturbation stochastic approximation) technique further improves the search results with effective initial solutions. This work enhances the practical values of tool path planning by providing a satisfactory machining quality.     

Supramolecular Glyco‐nanoparticles Toward Immunological Applications

Glyco‐mimicking nanoparticles (glyco‐NPs) with Förster resonance energy transfer (FRET) donor and acceptor groups formed via dynamic covalent bond of benzoboroxole and sugar from two complementary polymers are prepared. The glyco‐NPs are proved to be quite stable under physiological conditions but sensitive to pH. So the glyco‐NPs can be internalized by dendritic cells with integrity and nontoxicity and then dissociate within the acidic organelles. This particle dissociation is directly observed and visualized in vitro, for the first time via the FRET measurements and fluorescent microscopy. This feature makes controlled release of drug or protein by glyco‐NPs possible, i.e., when model antigen Ovalbumin is loaded in the glyco‐NPs, the released Ovalbumin in dendritic cells stimulates T cells more efficiently than the free Ovalbumin itself as a result of the enhanced antigen processing and presentation. Thus, the results enlighten a bright future of the glyco‐NPs in immunotherapy.     

基于Q-learning的不确定环境BDI Agent最优策略规划研究

BDI模型能够很好地解决在特定环境下的Agent的推理和决策问题,但在动态和不确定环境下缺少决策和学习的能力。强化学习解决了Agent在未知环境下的决策问题,却缺少BDI模型中的规则描述和逻辑推理。针对BDI在未知和动态环境下的策略规划问题,提出基于强化学习Q-learning算法来实现BDI Agent学习和规划的方法,并针对BDI的实现模型ASL的决策机制做出了改进,最后在ASL的仿真平台Jason上建立了迷宫的仿真,仿真实验表明,在加入Q-learning学习机制后的新的ASL系统中,Agent在不确定环境下依然可以完成任务。     

Model Selection for Unsupervised Learning of Visual Context

This study addresses the problem of choosing the most suitable probabilistic model selection criterion for unsupervised learning of visual context of a dynamic scene using mixture models. A rectified Bayesian Information Criterion (BICr) and a Completed Likelihood Akaike’s Information Criterion (CL-AIC) are formulated to estimate the optimal model order (complexity) for a given visual scene. Both criteria are designed to overcome poor model selection by existing popular criteria when the data sample size varies from small to large and the true mixture distribution kernel functions differ from the assumed ones. Extensive experiments on learning visual context for dynamic scene modelling are carried out to demonstrate the effectiveness of BICr and CL-AIC, compared to that of existing popular model selection criteria including BIC, AIC and Integrated Completed Likelihood (ICL). Our study suggests that for learning visual context using a mixture model, BICr is the most appropriate criterion given sparse data, while CL-AIC should be chosen given moderate or large data sample sizes.