在过去十年左右的时间里, 随着在无线传感网络, 群体机器人及无人飞行器编队等问题的广泛应用, 多智能体系统的协作控制问题已成为控制理论中的一个热点问题. 本论文将研究两类基本的协作控制问题: 多智能体趋同问题与多智能体协作式输出调节问题. / 作为多智能体系统协作控制的一个基本问题, 多智能体趋同问题是其它诸如蜂拥, 聚类, 编队等协作控制问题的基础. 目前趋同问题主要分为两类: 无领导者的趋同问题与有领导者的趋同问题. 无领导者的趋同问题的控制目标是设计分布式的控制器使得所有子系统的状态渐近地趋于一个共同但未知的轨迹, 而有领导者的趋同问题的控制目标是设计分布式的控制器使得所有子系统的状态渐近地趋于一个特殊的轨迹, 这个轨迹由一个特别的被称作领导者的子系统产生. 分布式的控制器常常由系统的拓扑连接图决定.连接图一般是时变的, 并包含固定图与切换图作为特例. 本论文的第一部分, 我们将研究连续时间一般线性多智能体系统与离散时间一般线性多智能体系统在联合连通假设下的切换拓扑网络的趋同问题. 该问题包含其它一些特殊的诸如一阶积分器系统, 筒谐振子系统的趋同问题作为特例. 这一部分的主要贡献概括为以下两点: / 1. 研究连续时间线性临界稳定多智能体系统在切换网络下的两类趋同问题. 为研究这两类问题, 我们将首先建立一类包含Kronecker 乘积的线性切换系统的稳定性结果. 该系统特别之处在于其系统矩阵在任何时刻都可以不是Hurwitz 稳定的. 我们将结合Lyapunov 稳定性理论与一类适用于分段连续线性系统的广义Barbalat 引理来研究该切换系统的稳定性. 作为该稳定性结果的直接应用, 我们分别给出求解两类趋同问题的静态状态反馈控制率. 与现有结果比较, 该结果仅假设动态图是联合连通的, 因而严格弱化了动态图的假设. / 2. 研究离散时间线性多智能体系统在切换网络下的两类趋同问题. 在系统矩阵是临界稳定的假设下, 我们证明如果动态图是联合连通的, 则都存在静态的分布式状态反馈控制器以达到两类趋同. 该趋同分析是基于一类自守线性离散时间切换系统的稳定性结果. 研究该切换系统的稳定性的主要困难在于系统矩阵在任何时刻都可以不是Schur 稳定的. 我们将结合共同Lyapunov 函数与一些新技巧来实现稳定性分析. 该结果将包含一些现有结果作为特例. / 论文的第二部分将研究多智能体系统的协作式输出调节问题. 该问题允许各子系统有不同的动态, 允许各子系统模型存在不确定性, 并且允许各子系统存在外部干扰, 因此该问题的描述较有领导者的趋同问题更为一般于实际. 该问题的控制目标是要利用分布式的控制策略来实现不确定多智能体系统的渐近跟踪和干扰抑制. 由于该问题描述的一般性, 其结论将包含其它一些诸如趋同, 同步, 编队等多智能体协同控制问题作为特例.就技术路线而言, 我们将建立分布式的观测器与分布式的内模来处理该问题. 这部分的主要贡献总结为以下三点: / 1. 研究线性多智能体系统分别在静态与切换拓扑网络下的协作式输出调节问题. 全系统包含两类子系统. 第一类子系统可以接收到外部系统的信号, 而第二类子系统不能接收到外部系统的信号. 因此, 传统的集中式的控制器与分散式的控制器都不适用于该系统. 我们将建立分布式的观测器实现外部系统的信息从第一类子系统向第二类子系统传递. 对静态拓扑网络情况, 我们分别给出分布式动态全状态反馈控制器与分布式动态测量输出反馈控制器求解该问题的充分必要条件. 对切换拓扑网络情况, 我们则给出分布式动态全状态反馈控制器与包含前馈项的分布式动态测量输出反馈控制器求解该问题的充分条件. 该结果可以作为多智能体有领导者的趋同问题的直接推广,并将应用于求解群体机器人有领导者的编队问题. / 2. 研究不确定线性多智能体系统在静态拓扑网络下的协作式鲁棒输出调节问题. 相对前一问题, 该问题允许多智能体系统的模型具有不确定参数. 因此前馈设计方案不适用于该问题. 通过建立分布式的内模, 我们将该问题转换成其增广系统的同时极点配置问题. 利用LQR 设计方法, 我们分别给出分布式动态状态反馈控制器与分布式输出反馈控制器求解该问题的充分必要条件. 由于极点配置具有鲁棒性, 这两类控制器均能容忍系统不确定参数的微小变化. 该结果也包含一些有领导者的趋同问题作为特例. / 3. 研究一类具有参数不确定性的混杂的多智能体系统在静态拓扑网络下的协作式鲁棒输出调节问题. 与前一问题比较, 这里我们允许系统参数在一个任意大的规定的紧集内变化. 为实现这一目标, 我们引入一类新的内模, 它能将协作式输出调节问题转化成其增广系统的鲁棒镇定问题. 我们将结合共同高增益状态反馈技巧与分布式高增益观测器技巧来设计分布式动态输出反馈控制器以求解该问题, 并同时给出其可解性的充分必要条件. 该结果可应用于求解一大类不确定多智能体系统的有领导者的鲁棒趋同问题. / Over the past decade, the extensive applications of wireless sensor networks, cooperative robotics, unmanned aerial vehicle formations and so on have made the cooperative control of the multi-agent system a trendy topic. This thesis will concentrate on two basic cooperative control problems: consensus and cooperative output regulation. / Consensus problem is one of the basic cooperative control problems of multi-agent systems. It is the foundation of many other cooperative control problems such as flocking, rendezvous, and formation control. There are two types of consensus problems: leaderless consensus problem and leader-following consensus problem. While the leaderless consensus problem aims to design a distributed controller for a multi-agent system so that the states of all agents asymptotically approach a common trajectory, leader-following consensus problem further requires that the distributed controller is such that the states of all agents converge to a specified trajectory which is usually produced by another agent called leader. The distributed controller is defined by a communication graph which is in general time-varying and contains both the fixed graph and switching graph as special cases. In the first part of this thesis, we will consider these two consensus problems for both continuous-time and discrete-time general linear multi-agent systems subject to the switching network topology under the jointly connected assumption. Our problem formulation includes the consensus of many typical physical multi-agent systems such as single-integrators and harmonic oscillators as special cases. The main results of this part are summarized as follows: / 1. Two consensus problems of continuous-time marginally stable linear multi-agent systems under switching network topology are studied. We first establish a stability result for a class of linear switched systems involving Kronecker product. The problem is intriguing in that the system matrix does not have to be Hurwitz at any time instant. We then establish the main stability result by a combination of the Lyapunov stability analysis and a generalized Barbalat’s Lemma applicable to piecewise continuous linear systems. As applications of this stability result, we present two distributed static state feedback controllers to solve the two consensus problems, respectively. In contrast with existing results, our result only assumes that the dynamic graph is jointly connected which is strictly weaker than any other assumptions. / 2. Two consensus problems of linear discrete-time multi-agent systems under switching network topology are studied. Under the assumption that the system matrix is marginally stable, we show that both leaderless consensus problem and leaderfollowing consensus problem can be achieved via the distributed static state feedback controllers provided that the dynamic graph is jointly connected. The consensus analysis is based on the stability analysis of a class of linear autonomous discretetime switched systems. The main difficulty to overcome is that the system matrix of such linear switched system may not be Schur at any time instant. We combine the common Lyapunov function approach with some novel technique to complete such stability analysis. Our result contains several existing results as special cases. / The second part of this thesis addresses the cooperative output regulation of linear multi-agent systems. The formulation of the cooperative output regulation problem is much more general than the leader-following consensus problem in that it deals with agents with different dynamics, allows model uncertainty, and accommodates external disturbance. The direct objective of this problem is to handle the asymptotic tracking and disturbance rejection problem in an uncertain multi-agent system via a distributed control approach. Due to the generality of this problem formulation, our result will also contain many control problems of multi-agent systems such as consensus, synchronization, and formation as special cases, thus leading to a unified solution to several different control problems of multi-agent systems. Technically, the distributed observer and the distributed internal model will be established for handling this problem. The main contributions of this part are summarized as follows: / 1. The cooperative output regulation of linear multi-agent systems under both static and switching communication network topologies is studied. The overall system consists of two groups of subsystems. While the first group of subsystems can access the exogenous signal, the second cannot. As a result, the problem cannot be solved by either the centralized approach or the decentralized approach. A distributed observer is devised so that it can relay the information of the exosystem from the first group to the second group. For the static network case, we present the sufficient and necessary solvability conditions via distributed dynamic state feedback control law and the distributed dynamic measurement output feedback control law. For the switching network case, we give the sufficient solvability condition via distributed dynamic state feedback control law and the distributed dynamic measurement output feedback with feedforward control law. This result can be viewed as a generalization of some leader-following consensus problems of multi-agent systems. It can also be applied to solve the leader-following formation problem of a group of mobil robots. / 2. The cooperative robust output regulation problem of linear uncertain multi-agent systems under static network topology is studied. In this problem, the structural plant uncertainty is further taken into consideration. Then the feedforward design is no longer applicable to this problem. By utilizing a distributed internal model, this problem is converted into a simultaneous eigenvalue placement problem of the so called augmented system. Using the LQR design method, we present the sufficient and necessary solvability conditions of this problem via both distributed dynamic state feedback control law and distributed dynamic output feedback control law. Due to the robustness of eigenvalue placement, such control laws can tolerate small plant uncertainty. This result also contains the leader-following consensus problem for several systems as special cases. / 3. The cooperative robust output regulation of a class of heterogeneous linear multiagent systems with parameter uncertainties under static network topology is studied. In contrast with the previous problem, here we allow the plant uncertain parameters to lie on an arbitrarily large prescribed compact subset. For this purpose, we introduce a new type of internal model that allows the cooperative robust output regulation problem of the given plant to be converted into a robust stabilization problem of an augmented multi-agent system. We then solve this problem via distributed dynamic output feedback control law by combining a simultaneous high gain state feedback control technique and a distributed high gain observer technique. The sufficient and necessary solvability conditions are also given. A special case of our result leads to the solution of the leader-following robust consensus problem for a large class of uncertain multi-agent systems. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Su, Youfeng. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2012. / Includes bibliographical references (leaves 155-164). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese. / Abstract --- p.i / Acknowledgement --- p.vi / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Literature Review --- p.1 / Chapter 1.2 --- Thesis Contributions --- p.5 / Chapter 1.3 --- Thesis Organization --- p.7 / Chapter 2 --- Preliminaries --- p.10 / Chapter 2.1 --- Review of Graph Notation --- p.10 / Chapter 2.2 --- Review of Linear Output Regulation --- p.12 / Chapter 3 --- Continuous-Time Consensus under Switching Network Topology --- p.20 / Chapter 3.1 --- Introduction --- p.20 / Chapter 3.2 --- A Stability Result --- p.22 / Chapter 3.3 --- Problem Statement --- p.29 / Chapter 3.4 --- Solvability of Two Continuous-Time Consensus Problems --- p.32 / Chapter 3.4.1 --- Leaderless Consensus --- p.32 / Chapter 3.4.2 --- Leader-Following Consensus --- p.35 / Chapter 3.5 --- Examples --- p.38 / Chapter 3.6 --- Conclusion --- p.43 / Chapter 4 --- Discrete-Time Consensus under Switching Network Topology --- p.44 / Chapter 4.1 --- Introduction --- p.44 / Chapter 4.2 --- Problem Statement --- p.45 / Chapter 4.3 --- A Stability Result --- p.48 / Chapter 4.4 --- Solvability of Two Discrete-Time Consensus Problems --- p.53 / Chapter 4.4.1 --- Leaderless Consensus --- p.53 / Chapter 4.4.2 --- Leader-Following Consensus --- p.55 / Chapter 4.5 --- Examples --- p.57 / Chapter 4.6 --- Conclusion --- p.64 / Chapter 5 --- Linear Cooperative Output Regulation under Static Network --- p.67 / Chapter 5.1 --- Introduction --- p.67 / Chapter 5.2 --- Problem Statement --- p.70 / Chapter 5.3 --- Solvability of the Problem --- p.71 / Chapter 5.3.1 --- Distributed State Feedback --- p.71 / Chapter 5.3.2 --- Distributed Measurement Output Feedback --- p.76 / Chapter 5.4 --- An Example --- p.82 / Chapter 5.5 --- Conclusion --- p.86 / Chapter 6 --- Linear Cooperative Output Regulation under Switching Network --- p.87 / Chapter 6.1 --- Introduction --- p.87 / Chapter 6.2 --- Problem Statement --- p.88 / Chapter 6.3 --- Solvability of the Problem --- p.90 / Chapter 6.3.1 --- Some Lemmas --- p.90 / Chapter 6.3.2 --- Distributed State Feedback --- p.96 / Chapter 6.3.3 --- Distributed Measurement Output Feedback with Feedforward --- p.97 / Chapter 6.4 --- Application to Leader-Following Consensus --- p.100 / Chapter 6.5 --- Two Examples --- p.103 / Chapter 6.6 --- Conclusion --- p.113 / Chapter 7 --- Linear Cooperative Robust Output Regulation: A Structurally Stable Approach --- p.114 / Chapter 7.1 --- Introduction --- p.114 / Chapter 7.2 --- Problem Statement --- p.116 / Chapter 7.3 --- Solvability of the Problem --- p.117 / Chapter 7.4 --- An Example --- p.123 / Chapter 7.5 --- Conclusion --- p.125 / Chapter 8 --- Cooperative Robust Output Regulation of Heterogeneous Linear Uncertain Multi-Agent Systems --- p.128 / Chapter 8.1 --- Introduction --- p.128 / Chapter 8.2 --- Problem Statement --- p.130 / Chapter 8.3 --- From Output Regulation to Stabilization --- p.131 / Chapter 8.4 --- Stabilization of the Augmented System --- p.134 / Chapter 8.4.1 --- Two Lemmas --- p.135 / Chapter 8.4.2 --- Stabilization via State Feedback --- p.137 / Chapter 8.4.3 --- Stabilization via Output Feedback --- p.140 / Chapter 8.5 --- Solvability of Cooperative Output Regulation --- p.142 / Chapter 8.6 --- Examples --- p.144 / Chapter 8.6.1 --- Leader-Following Tracking of Mass-Damper-Spring Systems --- p.145 / Chapter 8.6.2 --- Formation of Multi Vehicles with Unknown Amplitude Disturbance --- p.148 / Chapter 8.7 --- Conclusion --- p.151 / Chapter 9 --- Conclusions --- p.152 / Bibliography --- p.155 / Biography --- p.165

Recent advances in micro-electro-mechanical systems (MEMS) fabrication have made it possible to construct miniature devices containing an embedded system with strong computing capabilities. New generations of low cost sensor nodes can be created small with powerful computing and sensing capabilities. The small sensor nodes together with distributed wireless networking techniques enable the creation of innovative self-organized and peer-to-peer large scale wireless sensor networks (WSNs). A coordinated network of sensor nodes can perform distributed sensing of environmental phenomena over large-scale physical spaces and enable reliable monitoring and control in various applications. WSNs provide bridges between the virtual world of information technology and the real physical world. They represent a fundamental paradigm shift from traditional inter-human personal communications to autonomous inter-device communications. This thesis investigates the problems of target detection and tracking in WSNs. WSNs have some unique advantages over traditional sensor networks. However, the severe scarcity of power, communication and computation resources imposes some major challenges on the design and applications of distributed protocols for WSNs. In particular, this thesis focuses on two aspects of remote target detection and tracking in WSNs: distributed source coding (DSC) and sensor node localization. The primary purpose is to improve the application performance while minimizing energy consumption and bandwidth overhead.

"Statistical inference plays an increasingly important role in science, finance and industry. Despite the extensive research and wide application of statistical inference, most of the efforts focus on uniform models. This thesis considers the statistical inference in models with abrupt changes instead. The task is to estimate change-points where the underlying models change. We first study low dimensional linear regression problems for which the underlying model undergoes multiple changes. Our goal is to estimate the number and locations of change-points that segment available data into different regions, and further produce sparse and interpretable models for each region. To address challenges of the existing approaches and to produce interpretable models, we propose a sparse group Lasso (SGL) based approach for linear regression problems with change-points. Then we extend our method to high dimensional nonhomogeneous linear regression models. Under certain assumptions and using a properly chosen regularization parameter, we show several desirable properties of the method. We further extend our studies to generalized linear models (GLM) and prove similar results. In practice, change-points inference usually involves high dimensional data, hence it is prone to tackle for distributed learning with feature partitioning data, which implies each machine in the cluster stores a part of the features. One bottleneck for distributed learning is communication. For this implementation concern, we design communication efficient algorithm for feature partitioning data sets to speed up not only change-points inference but also other classes of machine learning problem including Lasso, support vector machine (SVM) and logistic regression."

A common problem when running Web-based applications is how to scale-up the database. The solution to this problem usually involves having a smart Database Administrator determine how to spread the database tables out amongst computers that will work in parallel. Laying out database tables across multiple machines so they can act together as a single efficient database is hard. Automated methods are needed to help eliminate the time required for database administrators to create optimal configurations. There are four operators that we consider that can create a search space of possible database layouts: 1) denormalizing, 2) horizontally partitioning, 3) vertically partitioning, and 4) fully replicating. Textbooks offer general advice that is useful for dealing with extreme cases - for instance you should fully replicate a table if the level of insert to selects is close to zero. But even this seemingly obvious statement is not necessarily one that will lead to a speed up once you take into account that some nodes might be a bottle neck. There can be complex interactions between the 4 different operators which make it even more difficult to predict what the best thing to do is. Instead of using best practices to do database layout, we need a system that collects empirical data on when these 4 different operators are effective. We have implemented a state based search technique to try different operators, and then we used the empirically measured data to see if any speed up occurred. We recognized that the costs of creating the physical database layout are potentially large, but it is necessary since we want to know the "Ground Truth" about what is effective and under what conditions. After creating a dataset where these four different operators have been applied to make different databases, we can employ machine learning to induce rules to help govern the physical design of the database across an arbitrary number of computer nodes. This learning process, in turn, would allow the database placement algorithm to get better over time as it trains over a set of examples. What this algorithm calls for is that it will try to learn 1) "What is a good database layout for a particular application given a query workload?" and 2) "Can this algorithm automatically improve itself in making recommendations by using machine learned rules to try to generalize when it makes sense to apply each of these operators?" There has been considerable research done in parallelizing databases where large amounts of data are shipped from one node to another to answer a single query. Sometimes the costs of shipping the data back and forth might be high, so in this work we assume that it might be more efficient to create a database layout where each query can be answered by a single node. To make this assumption requires that all the incoming query templates are known beforehand. This requirement can easily be satisfied in the case of a Web-based application due to the characteristic that users typically interact with the system through a web interface such as web forms. In this case, unseen queries are not necessarily answerable, without first possibly reconstructing the data on a single machine. Prior knowledge of these exact query templates allows us to select the best possible database table placements across multiple nodes. But in the case of trying to improve the efficiency of a Web-based application, a web site provider might feel that they are willing to suffer the inconvenience of not being able to answer an arbitrary query, if they are in turn provided with a system that runs more efficiently.

This thesis addresses the need for a scalable distributed solution for k-nearest-neighbor (kNN) search, a fundamental data mining task. This unsupervised method poses particular challenges on shared-nothing distributed architectures, where global information about the dataset is not available to individual machines. The distance to search for neighbors is not known a priori, and therefore a dynamic data partitioning strategy is required to guarantee that exact kNN can be found autonomously on each machine. Pivot-based partitioning has been shown to facilitate bounding of partitions, however state-of-the-art methods suffer from prohibitive data duplication (upwards of 20x the size of the dataset). In this work an innovative method for solving exact distributed kNN search called PkNN is presented. The key idea is to perform computation over several rounds, leveraging pivot-based data partitioning at each stage. Aggressive data-driven bounds limit communication costs, and a number of optimizations are designed for efficient computation. Experimental study on large real-world data (over 1 billion points) compares PkNN to the state-of-the-art distributed solution, demonstrating that the benefits of additional stages of computation in the PkNN method heavily outweigh the added I/O overhead. PkNN achieves a data duplication rate close to 1, significant speedup over previous solutions, and scales effectively in data cardinality and dimension. PkNN can facilitate distributed solutions to other unsupervised learning methods which rely on kNN search as a critical building block. As one example, a distributed framework for the Local Outlier Factor (LOF) algorithm is given. Testing on large real-world and synthetic data with varying characteristics measures the scalability of PkNN and the distributed LOF framework in data size and dimensionality.

"This thesis develops a system for synchronizing two wireless transmitters so that they are able to implement a distributed beamformer in several different channel models. This thesis considers a specific implementation of the system and proposes a metric to quantify its performance. The system's performance is investigated in single-path and multi-path time-invariant channel scenarios, as well as in single-path time-varying channel scenarios. Where prior systems have difficulty in implementing a distributed beamformer in multi-path channels and/or mobile scenarios, the results of this thesis show that the Round-Trip Time-Division distributed beamforming system is able to perform as a beamformer in all three of the channel models considered. "

Many modern web applications run on distributed cloud systems, which allows them to scale their resources to match performance requirements. Scaling of resources at industry scales, however, is a financially-expensive operation, and therefore one that should involve a business justification rooted in customer quality-of-service metrics over more commonly-used utilization metrics. Additionally, changing the resources available to such a system is non-instantaneous, and thus a reasonable effort should be made to predict system performance at varying resource allocations and at various expected workloads. Common performance monitoring solutions look at general metrics such as CPU utilization or available memory. These metrics are at best an indirect means of evaluating customer experience, and at worst may provide no information as to whether users of a commercial application are satisfied with the product they have paid for. Instead, the use of application-specific metrics that accurately reflect the experience of system users, combined with research into how these metrics are affected by various tunable parameters, allows a company to make accurate decisions as to the desired performance perceived by their users versus the costs associated with providing that level of performance. This thesis uses a real-world software-as-a-service product as a case study in the development of quality-of-service metrics and the use of those metrics to determine business cases and costing packages for customers. The product used for this work is Phoenix, a state-of-the-art social media aggregation and analytics software-as-a-service web platform developed by Echosec Systems, Ltd. The product will be tested under realworld conditions on cloud hardware with a minimal test harness to ensure a realistic depiction of live production conditions. / Graduate

This thesis evaluates a method for extracting architectural dependencies and performance measures from an evolving distributed software system. The research goal was to establish methods of determining potential scalability issues in a distributed software system as it is being iteratively developed. The research evaluated the use of industry available distributed tracing methods to extract performance measures and queuing network model parameters for common user activities. Additionally, a method was developed to trace and collect system operations the correspond to these user activities utilizing automated acceptance testing. Performance measure extraction was tested across several historical releases of a real-world distributed software system with this method. The trends in performance measures across releases correspond to several scalability issues identified in the production software system. / Graduate

Typescript, etc. / Digitized by Kansas Correctional Industries

A multi-agent system consists of a number of agents, which may include software agents, robots, or even humans, in some application environment. Multi-robot systems are increasingly being employed to complete jobs and missions in various fields including search and rescue, space and underwater exploration, support in healthcare facilities, surveillance and target tracking, product manufacturing, pick-up and delivery, and logistics. Multi-agent task allocation is a complex problem compounded by various constraints such as deadlines, agent capabilities, and communication delays. In high-stake real-time environments, such as rescue missions, it is difficult to predict in advance what the requirements of the mission will be, what resources will be available, and how to optimally employ such resources. Yet, a fast response and speedy execution are critical to the outcome. This thesis proposes distributed optimisation techniques to tackle the following questions: how to maximise the number of assigned tasks in time restricted environments with limited resources; how to reach consensus on an execution plan across many agents, within a reasonable time-frame; and how to maintain robustness and optimality when factors change, e.g. the number of agents changes. Three novel approaches are proposed to address each of these questions. A novel algorithm is proposed to reassign tasks and free resources that allow the completion of more tasks. The introduction of a rank-based system for conflict resolution is shown to reduce the time for the agents to reach consensus while maintaining equal number of allocations. Finally, this thesis proposes an adaptive data-driven algorithm to learn optimal strategies from experience in different scenarios, and to enable individual agents to adapt their strategy during execution. A simulated rescue scenario is used to demonstrate the performance of the proposed methods compared with existing baseline methods.