濒海隐密作业UUV环境同态预估与任务重构问题研究

In order to solve the common problems of navigation safety and autonomous behavior in littoral sea secretly for unmanned underwater vehicle (UUV), the method is approved to solve the difficulty of data analysis, modeling and simulation, the humanlike depended mode decision-making, and so on. Especially, the difficulty of reconfiguration and control is stressed without human in the loop, while researching the problem of environmental homomorphism forecasting, fuzzy association and deliberating reconfiguration, and so on. With the help of UUV prior knowledge database and acousto-optic perception means, the program could create the mechanism of littoral environment characteristics of UUV homomorphism forecast and the method of typical acousto-optic visual automatic identification. In the same time, the littoral secret UUV environment perception, threat estimation and local environment dynamic modeling method are also been improved. In other hand, the key technology of the abnormal or unknown modal behavior reconfiguration and humanlike association is improved too. The theory about UUV autonomous behavior decision-making is enhanced. It also would improve the safety and behavioral skills of UUV littoral secret operation. The program would explore the architecture of control system for UUV with human-not-in-loop, conquer the capacity of littoral secret operation for independent behavior of UUV technology. The research also brings up the cultivation of innovation team which stands at the international academic frontier and an ability of sustainable development is formed for UUV. An armament progress and operational application of UUV are expedited. The program should do some significant thinking on how to improve the standards of UUV autonomy and must help us to catch up with the advanced world levels in the field of UUV autonomy.

为解决水下无人航行器（UUV）濒海隐密作业的航海安全、自主行为等共性问题，本项目拟采用数据分析、建模仿真与类人依赖模决策等手段，针对人不在环的控制与重构难点，开展环境同态预估、模糊匹配与慎思重构等问题研究。利用UUV先验知识和声光视觉等感知手段，创建濒海环境特征的UUV同态预估机理及声光视觉典型特征自动识别方法；提出濒海隐密作业UUV的环境感知、威胁估计与局部环境动态建模方法；突破异常或未知模态下的行为重构与模糊匹配等关键技术，完善UUV的自主行为研究理论体系，提升UUV濒海隐密作业行为技能及安全性。通过研究，探求UUV人不在环的自主控制理论体系框架，攻克濒海隐密作业UUV的自主行为能力的瓶颈技术，引领能傲立国际学术前沿创新团队的成长，促进UUV业务化应用广度与深度，奠定UUV自主控制理论基础，形成其可持续发展的潜力。

UUV主要依靠声学和光学设备感知与探测环境，获得信息呈现弱信号特点。人在回路有助于信号处理后的置信度提高。但UUV执行隐密作业时，需实时对环境感知、评估与决策，确保航海与作业时遭遇非预期事件的任务重构能力。面向弱观测、人不在环的特点，即时环境同态预估与任务重构等问题研究已成为关键和瓶颈问题。. 本项目为解决UUV濒海隐密作业的航海安全、自主行为等共性问题，围绕UUV的自主感知、自主管理、自主规划、自主决策等自主能力开展研究。首先，总结形成了UUV行为的先验规则库，并形成了形式化数学描述方法，为UUV具备在线自主规划与行为决策能力奠定了基础。在UUV声光视觉自主感知方面，提出了水下环境声光背景噪声关联检测的阈值自适应调节方法及水下目标特征的提取方法，为提高UUV对水下目标的探测和利用效果提供了有效手段。在海洋环境感知及应用方面，重点针对典型海流进行建模并分析了其对UUV的影响，提出了UUV对局部流场信息的关联估计方法，实现了利用模型先验知识与测量数据拟合对海流的连续时间关联估计。提出了基于环境关联的自主导航同态预报方法，为UUV自适应利用环境提升远程自主航海能力奠定了基础。在故障及异常行为安全响应管理方面，总结提出了UUV典型故障模式及故障特征、UUV异常行为模式及安全响应机制，为UUV具备自身故障及功能退化下的安全响应能力提供了重要参考。在面临障碍环境的自主行为决策方面，建模了濒海环境下影响UUV 航行的碍航物特征，提出了复杂障碍场景下的UUV深度学习算法和机理，可以实现UUV在复杂障碍环境下的安全航行和作业。在任务自主规划及重构方面，提出了UUV示教样本集深度学习的动态规划方法、多属性慎思的容错规划与调度方法以及非预期遭遇威胁评估及任务重构方法，可以大幅提升UUV对复杂多任务的规划能力以及人物执行过程中自主重规划能力。