毕业生知识库

    移动机器人的自主环境探索和主动定位是其在未知环境下开展任务的前提和关键，因此具有重要的理论研究意义和广泛的应用前景。本文针对移动机器人在未知环境下的探索路径规划与自主定位方法等方面展开了研究。
    论文的主要内容如下：

    一、针对移动机器人在室内未知环境下的探索路径规划问题，提出一种基于信息熵地图的探索路径规划方法。环境信息熵地图不但能够反映环境未知区域的探索情况，而且还可有效评估规划路径的信息熵增益，一定程度上解决了移动机器人在未知环境下高效、全局覆盖探索路径的规划问题，实现了移动机器在未知环境下快速全面的环境探索。实验结果验证了所提方法的有效性。

    二、针对移动机器人在环境定位时遇到的动态目标干扰问题，提出一种动态目标扰动下的主动视觉定位方法。该方法实现了基于目标检测和运动状态估计的场景动态目标识别方法，同时动态目标的识别结果也被加入到相机观测姿态的优化与调整之中，主动减弱甚至消除了动态目标对视觉定位带来的扰动影响。实验结果表明所提方法能够很好地克服动态目标扰动实现主动视觉定位。

    三、针对视觉定位中因环境纹理缺失、相机视角遮挡以及快速运动等导致的定位易于丢失的问题，提出一种基于环境特征分布反馈的相机观测姿态优化和调整方法。该方法能够根据实时获取的环境特征分布情况，自动优化并调整相机的观测姿态，在一定程度上解决了上述异常情况下视觉定位易于丢失的问题。仿真和实验的结果都验证了所提方法的有效性。

    四、针对室内移动机器人在初始化或运行过程中视觉定位丢失后的重定位问题，提出一种基于疆界探索的移动机器人自主重定位方法，通过引导机器人对局部特征地图进行快速甚至是完整的遍历，从而实现了自主的重定位。仿真和实验的结果都验证了所提方法的有效性。

    最后，总结了本文所取得的成果，并对下一步可以开展的研究做了展望。

Autonomous environment exploration and active positioning capability for mobile robots is the basis for mobile robots to execute tasks autonomously, and it is crucial to execute tasks effectively. It is significant in both research and applications. This thesis focuses on the environmental exploration path planning and active visual positioning method for mobile robots. The main contents of this thesis are as follows.

Firstly, to address the path planning for mobile robots exploring autonomously in an indoor environment, an exploration path planning method is proposed based on an environmental information entropy map. The constructed environment information entropy map can not only reflect the distribution of unknown regions in the process of exploration but also assess information gain on the planned path. As a result, mobile robots can generate the most efficient environment exploration path, and explore rapidly the environment. Experimental results verify the effectiveness of the proposed method.

Secondly, due to dynamic target interference encountered by mobile robots during environmental positioning, an autonomous visual positioning method is given under dynamic target disturbance. The proposed method can detect and track the target in the scene, and estimate the motion state of the target. Based on the obtained estimation of the motion state, the dynamic target in the scene can be correctly found.

Thirdly, to solve the problem of positioning loss caused by uneven distribution of environmental textures, missing and rapid camera movement encountered in visual positioning, a method is proposed for optimizing and adjusting the camera observation posture based on the feedback of environmental feature distribution. The proposed method can automatically optimize and adjust the observation posture of the camera based on the feedback of the environmental feature distribution obtained from the camera. It alleviates the impact of problems such as image quality degradation caused by different environmental characteristics and rapid motion on visual positioning. Simulation and experiment results illustrate the performance of the proposed method.

Fourthly, aiming at the initial relocalization of mobile robots and relocalization after positioning loss during task execution, a method for autonomous relocation of mobile robots based on boundary exploration is proposed. This method can establish a local feature map based on the previous location information, and adopt a feature backtracking strategy based on the frontier exploration to achieve rapid and autonomous relocalization for mobile robots.

Finally, the conclusions are given and the future work is addressed.