Coactive design of explainable agent-based task planning and deep reinforcement learning for human-UAVs teamwork

Unmanned Aerial Vehicles (UAVs) are useful in dangerous and dynamic tasks such as search-and-rescue, forest surveillance, and anti-terrorist operations. These tasks can be solved better through the collaboration of multiple UAVs under human supervision. However, it is still difficult for human to monitor, understand, predict and control the behaviors of the UAVs due to the task complexity as well as the black-box machine learning and planning algorithms being used. In this paper, the coactive design method is adopted to analyze the cognitive capabilities required for the tasks and design the interdependencies among the heterogeneous teammates of UAVs or human for coherent collaboration. Then, an agent-based task planner is proposed to automatically decompose a complex task into a sequence of explainable subtasks under constrains of resources, execution time, social rules and costs. Besides, a deep reinforcement learning approach is designed for the UAVs to learn optimal policies of a flocking behavior and a path planner that are easy for the human operator to understand and control. Finally, a mixed-initiative action selection mechanism is used to evaluate the learned policies as well as the human’s decisions. Experimental results demonstrate the effectiveness of the proposed methods.     

Minimizing human-machine interface failures in high risk systems

Technology now permits the building of very complex man-machine systems with centralized controls, with the result that many processes can be run by relatively few individual workers. Studies of failures within these complex systems indicate that they are usually the consequence of a series of highly complex coincidences. There is an institutional neglect or misunderstanding of the implications of low-probability, high-consequence events for the design of complex man-machine systems. We must stop designing systems in which we virtually guarantee that operator errors will occur with catastrophic consequences. The greatest payback in reducing high risk system accidents is to reduce catastrophes induced or exacerbated by human error. This paper discusses some task breakdowns between the human element and software/hardware. These task allocations allow the complex man-machine system to be designed more robustly and prevent human error so as to reduce possible catastrophic consequences     

预防性维修任务可执行区间权衡确定方法研究

直接维修成本（DMC）是航空器制造商向运营人进行经济性担保的一项主要参数,是运营阶段监控维修成本的参考比较基准。提出一种兼顾考虑DMC 担保值和计划维修要求所列维修间隔的预防性维修任务可执行区间权衡确定方法,分析该方法的数据采集需求,介绍其使用流程,对度量预防性维修任务经济性的指标参数进行定义,进而给出维修任务可执行区间的确定方法。结果表明：该方法能够从经济可承受性的角度确定出每一项预防性维修任务的可执行区间,从而为航空器运营人整合实施维修任务提供参考,达到控制优化维修成本的目的。     

Operator scheduling strategies in supervisory control of multiple UAVs

The application of network centric operations to time-constrained command and control environments will mean that human operators will be increasingly responsible for multiple simultaneous supervisory control tasks. One such futuristic application will be the control of multiple unmanned aerial vehicles (UAVs) by a single operator. To achieve such performance in complex, time critical, and high risk settings, automated systems will be required both to guarantee rapid system response, as well as manageable workload for operators. Through the development of a simulation test bed for human supervisory control of multiple independent UAVs by a single operator, this paper presents recent efforts to investigate workload mitigation strategies as a function of increasing automation. A human-in-the-loop experiment revealed that under low workload conditions, operators' cognitive strategies were relatively robust across increasing levels of automated decision support. However, when provided with explicit automated recommendations and with the ability to negotiate with external agencies for delays in arrival times for targets, operators inappropriately fixated on the need to globally optimize their schedules. In addition, without explicit visual representation of uncertainty, operators tended to treated all probabilities uniformly. This study also revealed that operators who reached cognitive saturation adapted two very distinct management strategies, which led to varying degrees of success. Lastly, operators with management-by-exception decision support exhibited evidence of automation bias.     

有人机/无人机协同任务控制系统

针对有人机/无人机协同任务控制系统的设计和实现问题展开研究。结合有人机/无人机协同控制过程的特点,提出了基于自然语言接口方式的有人机/无人机协同任务控制系统结构。在此基础上,分析了系统模块的关键内容,设计了面向有人机与无人机交互过程的任务指令格式,提出了基于优先级的任务模式调度策略,针对无人机的典型任务模式,讨论了航路规划的计算方法选择,最后构建了有人机/无人机协同任务仿真环境。仿真试验结果验证了整体方法的可行性。     

基于CNS性能及人为因素的纵向碰撞风险研究

为了分析CNS性能及人为因素对航空器在平行航路的安全间距影响,基于雷达管制环境下管制员对航空器纵向穿越的干预过程,在传统碰撞风险模型的基础上引入认知可靠性与失误分析方法(CREAM),综合考虑通信、导航、监视(Communication,Navigation and Surveillance,CNS)性能以及人为因素,建立了包含人的认知可靠性的平行航路纵向碰撞风险模型.并通过算例分析了CNS性能以及人因可靠性对平行航路碰撞风险的影响.结果表明,导航性能对航路的纵向碰撞风险影响较大;通过提高人的认知可靠性可以有效地提高平行航路的安全水平.     

飞行系统中人的可靠性评估

人的可靠性是个非常活跃的研究领域。本文从航空事件的历史数据出发,论证了飞行系统中人员失误率评估的重要性和迫切性。综述了近年来国内外人的可靠性分析的方法和模型,它们是HCR,CSE,NRC,TESEO,SLIM和灰色模型;评述了各种模型的适用条件及优劣,特别是文中提出的灰色模型将是人员失误率评估中最为有效的方法。同时也为飞行系统的系统配置、座舱资源管理等提出了新的设计考虑因素。     

多无人机监控系统人机交互界面研究

首先从人机权限方面考虑,提出基于三级自主等级以上的无人机系统,分析典型无人机作战任务下各操作项自动化程度、复杂性等信息,得出多机监控能力下操作员角色需向指挥员转变。然后以四机监控为例,通过设计人机智能认知信息架构,建立操作员意图识别模型,采用空域页面布局切换和时域页面智能推显等形 式进行人机交互界面设计,实现单操作员对多架无人机的监控。最后在飞行模拟器环境下进行了多机监控仿真测试,测试结果证明本文设计的人机交互界面合理可行。     

基于CPN的飞机驾驶操纵过程建模及工效分析

驾驶舱人机界面的设计直接关系到飞行员在飞行任务中的操纵表现和飞行安全.分析了飞机驾驶操纵过程中的人机交互特性,基于着色Petri网构建了飞机操纵过程模型并给出了模型的公理化定义.提出了任务可达性,飞行员认知负荷和基于熵值的操纵程序复杂性评估方法,能够实现在驾驶舱设计初期对飞行员认知负荷与操纵过程的分析.结合某型飞机的起飞中断过程给出了操纵过程建模与工效分析实例,表明模型与方法的有效性.     

柏拉图理念论及其在西方认识发展史中的地位

柏拉图理念论作为哲学与神学的交汇点，形式是神学的，所阐述的内容却并非都是消极的。它以特定的方式论述了一般、共相是认识的唯一对象，以极端的方式肯定了主体认识结构在认识中的作用，以颠倒的形式阐述了认识是一个由感性上升到理性的辩证过程。这一学说对认识对象、认识起源以及认识过程所做的深刻揭示，对以后的西方认识发展史产生了深远影响。     

基于遗传算法的凸多边形区域航迹规划

提出了基于遗传算法的凸多边形区域航迹规划方法,针对飞行器转弯角和边界限制的航迹约束条件,改进了航迹编码方式,采用了基于方向编码的回溯方法生成航迹;同时改进了遗传算法的交叉算子和变异算子。仿真结果表明,该方法能快速有效地完成飞行器航迹规划任务,获得符合适应度要求的优化航迹。     

多轴多回路人机控制系统模型与仿真

针对驾驶员完成飞机机场下滑过程的多轴多回路控制任务,分析了人的控制行为特点,确定出驾驶员的信息反馈,形成了机系统的闭环控制结构,根据驾驶员对各控制回路的注意力分配和人机自适应规律性,选择相应的驾驶员模型形式并确定其参数,建立了多轴多回路的人机控制系统模型,最后,对风切变大气环境下驾驶员控制的飞机进场下滑过程进行了仿真计算,结果表明,所建立的人机控制系统模型是合理的,对飞行品质研究有参考价值。     

基于经验移植的自主空战对抗学习方法

现有的机器学习方法大多是交互式的学习模式，这类方法在训练过程非常依赖与环境之间的交互数据。空战对抗任务是一种奖励非常稀疏的训练任务，智能体在学习开始的很长一段时间内，都在探索能够获得奖励的动作。如果每一个新的任务都重新训练，是非常浪费计算资源的。因此，设计了一种基于经验移植的学习方法，使得经过训练的智能体能够将知识分享给新的智能体，提高其在新任务上的学习效率。首先，借鉴人类通过经验进行快速学习的现象，构建了基于经验移植的学习的模型；其次，兼顾知识分享和新任务的特征，明确了经验的内涵，建立了"知识+任务→经验"的融合认知方式；再次，设计了借鉴学习方法，将外部经验与任务相结合，进而转化为新个体的知识；最后，使用经验适用度作为筛选指标，分析了经验适用度对借鉴学习效率的影响，确定了执行借鉴学习的筛选边界。新个体通过借鉴学习后能够获得关于新任务的初步知识，在新任务中更快地找到能够获得奖励的动作策略，从而提升在新的任务中的学习速度。     

基于脉冲交错的数字阵列雷达任务优化调度算法

针对数字阵列雷达搜索、跟踪和成像任务的资源调度问题,提出一种数字阵列雷达(DAR)任务的优化调度算法。该算法以脉冲交错技术为基础,在对目标搜索与跟踪的同时,利用基于压缩感知的稀疏孔径认知逆合成孔径雷达(ISAR)成像方法对部分精密跟踪目标成像,并采用观测时间动态调整策略以提高雷达系统的自适应能力。仿真结果表明,与传统雷达资源调度算法相比,该算法可以将成像任务考虑到优化调度模型中并合理分配资源,实现雷达多任务并行的调度,获得更高的资源利用率与期望的成像质量。     

知识作业的内涵辨析与界定

借鉴泰勒对工作（task）的定义，结合现代认知科学信息加工、问题解决等理论，在分析比较知识作业与操作作业、知识作业与脑力劳动以及知识作业与问题解决的基础上，适应知识经济时代生产率提高的要求，从管理工程与工效、信息加工、问题解决三个角度给出知识作业的内涵，分别为：知识作业是“一连串的动作”，心智操作是知识作业过程的“动素”；知识作业是人脑的信息加工和计算过程，心智操作是知识作业的“算法”；知识作业是根据作业目标所确定的心智操作序列，包括结果的规范化输出。     

Multistage integration model for human egomotion perception

Human computational vision models that attempt to account for the dynamic perception of egomotion and relative depth typically assume a common three-stage process: first, compute the optical flow field based on the dynamically changing image; second, estimate the egomotion states based on the flow; and third, estimate the relative depth/shape based on the egomotion states and possibly on a model of the viewed surface. We propose a model more in line with recent work in human vision, employing multistage integration. Here the dynamic image is first processed to generate spatial and temporal image gradients that drive a mutually interconnected state estimator and depth/shape estimator. The state estimator uses the image gradient information in combination with a depth/shape estimate of the viewed surface and an assumed model of the viewer's dynamics to generate current state estimates; in tandem, the depth/shape estimator uses the image gradient information in combination with the viewer's state estimate and assumed shape model to generate current depth/shape estimates. In this paper, we describe the model and compare model predictions with empirical data.     

Impact and the design of the human-machine interface

In this paper, we consider the concept of the impact of an action or human error. We begin from an informal definition of impact as: the effect that an action or sequence of actions has on the safe and successful operation of a system; and develop a quantitative measure of the impact of specified behaviours. It is important that human-machine interface designers should understand the relationship between operator actions and the hazards associated with a system. We demonstrate how impact can be assessed prior to, or in parallel with, the design of the human-machine interface, and show how impact assessments could be used to allow risk analysts to inform designers about the relationship between operator actions and system hazards. To illustrate our approach we present a simple case study,     

基于改进烟花算法的资源分配

资源分配问题作为一个NP-Hard问题,在云计算、无线电、卫星调度、多无人机协同作业等领域皆有研究需求,是一个共性的数学问题。烟花算法作为一种智能优化算法,具有求解大规模资源分配问题的能力,但也存在求解精度低等问题。为了提高传统烟花算法的计算效率和全局寻优能力,提出一种改进烟花算法,用遗传算法中的变异算子替代高斯变异操作,并增加模拟退火流程。最后在多无人机协同作业任务分配数学模型上进行仿真验证,实验结果表明在收敛速度以及计算精度方面,该算法均优于其余3种烟花算法。     

What avionics engineers should know about pilots and automation

There has been great interest lately in automation issues, spurred partly by accidents and incidents attributed to automation factors, and partly by a recognition that automation related problems can be both highly subtle and very dangerous. One of the reasons underlying automation-related problems is the relatively indiscriminate application of automation based on technological availability rather than operator need. This has placed system operators in roles that are poorly suited to human capabilities. A more productive approach to automation and humans would be to recognize the need to maintain appropriate roles for the human operator, the need to match automation functionality to operator tasks, the need for the human interface to make automation functions highly visible to the operator, and the potential for operator over- and under-reliance on the automation     

基于多机协作的认知无人机网络能效联合优化

针对无人机（UAV）通信网络中频谱资源紧缺的问题,构建基于认知无线电的多无人机通信网络,通过多机协作频谱感知有效探索授权频谱。提出一种基于Bisection算法的迭代算法,通过联合优化感知时间和判决门限对构建的复杂非凸问题求解,显著提高了无人机次级认知网络的能量效率（EE）。分析了无人机飞行过程中能效的变化情况,仿真结果表明,存在最优感知时间使能效获得最大值,且判决门限的选择会影响该能效最优值;提出的高能效迭代算法具有较好收敛性,有效提高了认知无人机网络的能量利用率。