几何/拓扑混合空间中的仿生导航方法研究

项目名称： 几何/拓扑混合空间中的仿生导航方法研究

项目编号： No.61503403

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 张礼廉

作者单位： 中国人民解放军国防科技大学

项目金额： 22万元

中文摘要： 无人武器装备已成为国家间军事博弈的重要力量和新的常规威慑手段。目前的无人系统主要依赖于卫星导航技术和无线电遥控技术实现几何空间的导航定位。在“导航战”的挑战下，这些导航方式面临着卫星信号和无线电遥控信号被电磁干扰或欺骗而导航失效的巨大风险。本项目以仿生学为基础，通过模拟动物的导航机制，综合应用偏振光、视觉以及微惯性传感器，研究适用于无人系统的几何/拓扑混合空间中的仿生自主导航方法：(1)为了实现滤波器恒速更新，设计了基于三视图几何的惯性/视觉/偏振光导航算法；(2)为了提高复杂环境中场景识别的鲁棒性，设计了基于话题的场景概率描述方法；(3)为了克服场景识别中的多重匹配假设问题，设计了基于神经元模型的混合空间导航方法。本课题的研究为大范围、长航时的无人系统导航提供新的理论和技术支撑，能有效提高无人武器装备的战场生存能力和环境适应能力。

中文关键词： 多传感器融合；运动估计；场境识别；混合空间；仿生导航

英文摘要： Unmanned weaponry, as a new kind of conventional deterrence, has played a vital role in the military games. Currently, the navigation systems of unmanned vehicles mainly depend on the satellite navigation technology and wireless remote control technology for positioning and piloting. When facing the challenge of NAVWAR, these navigation technologies may fail because satellite signals and radio signals are easy to be disturbed or deceived.Based on the navigation strategies of nature creatures, this project aims to solve the autonomous navigation problem for unmanned vehicles in hybrid metric/topological space by combining polarization light sensor, visual sensor and MIMU. We pay our focus on the following three key components: (1)To achieve constant update time, a multi-sensor integration algorithm based on three-view geometry is proposed; (2)To improve the robustness of place recognition process, a topic-based probabilistic representation for scenes is designed; (3)To solve the multi-hypothesis problem in place recognition, the information fusion algorithm in hybrid metric/topological space based on the hippocampal neuron model is investigated.The achievements in this project will provide both theoretical and practical supports for unmanned vehicles in large-scale navigation and long-term navigation situations. The proposed technology will sufficiently improve the viability and adaptability of unmanned weaponry.

英文关键词： Multi-Sensor Fusion;Motion Estimation;Scene Recognition;Hybrid Space;Bio-inspired Navigation