仿动物空间认知的多机器人协作目标搜索方法研究

The cooperative goal foraging of multi-robots in unknown and dynamic environments is a critical and challenging issue in the robotic field. The key contents include the Simultaneous Location and Mapping algorithm (SLAM) and the cooperation approach of multi-robots. Most of the SLAM algorithms are under varying assumptions, such as the environment is static and the robot is equipped with accurate sensors and a high performance computer. Moreover, much of the research on the SLAM algorithms and the goal foraging problems is independent, so it is difficult to use the results obtained from the study of SLAM algorithm, for the reality applications such as goal foraging. However, as we know, animals appear to have a great ability in mapping, goal foraging and navigation, without high precision sensors and complex computation. Thus, to study and create a biologically inspired and efficient multi-robot cooperative goal foraging approach with practical performance is very significant. .This project attempts to use the bioinspired neural network to realize the simultaneous location and mapping, inspired from the animal spatial cognition. The proposed approach integrates the map expression, the external stimulation, the experience learning, the goal foraging and robot navigation together. The main research contents of this project are as follows: a SLAM algorithm based on the bioinspired neural network; a goal location and robot navigation approach based on the memory network and a multi-robot cooperation approach based on the experiences reinforced from other robots in the system. The successful implementation of this project will have broad application prospects in a lot of fields, such as the unknown environment exploration, the disaster rescue, the anti-terrorism and the battlefield search.

未知动态环境下多机器人目标搜索是机器人研究领域一个很重要又具挑战性的课题，其核心内容包括机器人同时定位与地图创建（SLAM）和多机器人协作。现有的SLAM算法大多假设环境是静态的，机器人有高精度传感器和高性能计算机，且很多SLAM和目标搜索问题的研究是割裂开的，无法充分利用SLAM研究的成果。动物具有很强的定位、导航和目标搜索能力，且不需要高精度传感器和复杂的计算，研究一种仿动物空间认知的、高效的、可用于实际的多机器人目标搜索方法具有重要意义。.本项目利用生物刺激神经网络，结合动物空间认知机制，进行机器人同时定位和地图创建，将地图表示、外界刺激、经验学习、目标搜索、导航等融为一体，主要研究内容包括基于生物刺激神经网络的SLAM算法、基于记忆网络的目标定位和机器人导航以及利用同伴经验强化的多机器人协作方法等。本项目的成功实施在未知环境探索、灾害救援、反恐、战场搜索等方面具有广阔的应用前景。

未知动态环境下的多机器人协作目标搜索是机器人研究领域一个很重要又具挑战性的课题，其关键技术包括两个方面，一个是未知环境下的机器人同时定位和地图创建（SLAM），二是多机器人协作目标搜索。目前的SLAM算法主要致力于实现有效的机器人导航和地图创建方法，但大多数方法都假设环境是静态的、机器人测距传感器是精确的，等等，且很多SLAM算法和目标搜索问题的研究是割裂开的，无法充分利用SLAM研究的成果。在多机器人协作方面，如何提高协作方法的效率和适应性是关键。动物具有很强的定位、导航和目标搜索能力，且不需要高精度传感器和复杂的计算，研究一种仿动物空间认知的、高效的、可用于实际的多机器人协作目标搜索方法具有重要意义。. 本研究结合动物空间认知机制，开展机器人在未知环境下协作目标搜索有关理论研究，包括SLAM算法、机器人路径规划与导航、多机器人编队控制等，本研究在生物刺激神经网络SLAM算法、基于位置细胞机器人定位、机器人路径规划和导航等方面取得较大研究进展。相关研究成果将为未知环境下多机器人协作目标搜索提供理论基础，并通过开展相关实验研究，为将来的实际应用提供技术支持，研究成果在未知环境探索、灾害救援、反恐、战场搜索与救援等方面具有广阔的应用前景。

内容获取失败，请点击重试