煤矿探测机器人导航关键技术研究

发布时间：2018-05-09 03:42

本文选题：探测机器人 + 定位　；参考：《中国矿业大学》2014年博士论文

【摘要】：本文根据煤矿井下尤其是灾后的非结构化的地形环境，结合国内外煤矿探测机器人导航的研究现状，对煤矿探测机器人导航所涉及的方法与技术进行研究。论文对煤矿探测机器人的导航环境进行分析，并以此为依据对煤矿探测机器人的运动装置及控制系统进行分析与设计实现，探讨了机器人定位、导航环境检测及识别方法，并对基于仲裁融合协调策略的导航方法进行了详细研究，得到了对提高煤矿探测机器人导航能力具有借鉴意义的结论。本文首先从煤矿井下正常巷道、工作面内的地形特征、设备布置的分析出发，研究机器人需要克服的煤矿井下环境，对正常生产与灾后的煤矿的工作面及巷道内的环境进行分类与特征提取，将煤矿探测机器人导航环境归纳为四种典型的地形，包括凸台、斜坡、连续阶梯和壕沟，并给出其尺寸范围。为了提高煤矿探测机器人的越障能力，设计一种基于非对称双三角形的履带悬架，并建立相应的理论模型，对其越障特性进行了分析。通过分析表明，该型摇杆式非对称双三角形履带机器人运动装置具有良好地被动地形适应能力与越障性能，可爬上0.32m高的阶梯，爬下0.52m高的阶梯，通过0.5m宽的壕沟。讨论了探测机器人控制系统的组成，并在系统功能分析的基础上对控制系统整体方案进行了设计，通过分析比较确定使用PSoC集成开发平台作为处理器，进一步根据原型机详细设计了控制系统的硬件电路和软件程序，并完成了PCB和实物的制作。并对机器人的越障性能进行试验验证，为导航策略规划提供参数依据。分析了经典卡尔曼滤波、衰减记忆滤波、限定记忆滤波的基本原理和它们的组合使用方式，在此基础之上，提出了带有指数渐消因子的限定记忆卡尔曼滤波方法作为煤矿救灾机器人的多传感器信息融合方法。根据煤矿救灾机器人的自主定位系统的构成，对基于航迹推算的自主定位方法进行了介绍，提出改进后的航迹推算方法，使用带有指数渐消因子的限定记忆卡尔曼滤波方法以逐级滤波的方式融合多传感器信息，实现了对里程计误差的最小化控制。研制了以STC12C5A60S2微控制器为核心的多传感器信息融合的硬件系统，并给出了软件算法流程。计算机仿真结果表明，滤波方法可以稳定地实现0.5m的跟踪精度。在此基础上对机器人的自主定位功能进行试验，验证了所用方法的有效性。然后，通过对几种匹配方法的研究，，针对机器人实时性的要求，提出了一种基于最优种子排序的准稠密匹配方法；讨论了利用对极线的单点法获取障碍物三维信息的方法，首先分割障碍物边缘，利用对极线进行加速匹配，逐个求取障碍物各个边缘极大点的匹配点，进而求出各个边缘极大点的摄像机坐标，然后求出障碍物三维信息——高度、宽度和距离；建立了坡度计算的数学模型，并利用RBF方法得到在非正对斜坡情况下对坡度地形进行估算的模型；采用Canny算法对连续阶梯图像边缘进行了检测，应用最小二乘法进行拟合直线处理，给出了阶梯偏移角的测算方法。接着，为了提高煤矿探测机器人对导航环境的理解能力，根据本文设计的CUMT-5型煤矿探测机器人的越障能力，同时利用模糊理论、图像处理技术和神经网络技术等手段，对地形的可通过性的分析和评价方法进行了探讨，提出了对空间位置信息和视觉传感信息融合从地形起伏度、地形开阔度、地形坡度、地形不连续度和连续阶梯影响度五个方面对导航环境可通过性进行评价的方法。在对模糊理论、职能学习和神经网络技术等手段进行融合的基础上，创造性地提出了一种基于仲裁融合协调策略的行为控制导航方法。并进一步基于典型的环境模型对行为控制器进行设计，将融合机制和仲裁机制的优点融入到行为协调策略中，实现了自适应地对各基本行为的协调控制。在此基础上，将地形可通过性评价引入到协调策略中，进一步实现了煤矿探测机器人在复杂非结构环境中的导航。最后，对机器人综合导航进行对比实验，结果表明本章提出的导航方法能够实现复杂非结构化环境下的自主导航，不仅对不同环境具有较好的适应能力和可靠性，而且还具有较高的能量约束能力。
[Abstract]:According to the unstructured terrain environment in coal mine, especially the unstructured terrain environment after the disaster, combined with the research status of the navigation of the coal mine detection robot at home and abroad, this paper studies the methods and techniques involved in the navigation of the coal mine detection robot. The paper analyzes the navigation environment of the coal mine detection robot, and based on this, it is based on the coal mine detection robot. The analysis and design of the motion device and control system are carried out. The methods of robot positioning, navigation environment detection and recognition are discussed, and the navigation method based on the coordination strategy of arbitration is studied in detail, and the conclusions are obtained for improving the navigation ability of the coal mine detection robot.
In this paper, starting from the normal coal mine tunnel, the terrain features in the working face and the analysis of the equipment layout, this paper studies the underground environment of the coal mine which the robot needs to overcome, classifies and extracts the environment in the working face and the roadway of the coal mine after the normal production and after the disaster, and sums up the navigation environment of the coal mine detection robot into four typical types. Terrain, including platform, slope, continuous staircase and trench, and its size range.
In order to improve the obstacle avoidance ability of the coal mine detection robot, a caterpillar suspension based on asymmetric double triangle is designed and the corresponding theoretical model is set up to analyze its barrier characteristics. Through analysis, it is shown that the rocker type asymmetrical double triangular crawler robot has a good adaptability and obstacle avoidance of the passive terrain. Performance, can climb up the 0.32m high ladder, climb down the 0.52M high ladder, through the 0.5m wide trench. Discuss the composition of the control system of the detection robot, and design the whole scheme of the control system on the basis of the function analysis of the system. Through the analysis and comparison, the PSoC set development platform is used as the processor, and the prototype is further based on the prototype. The hardware and software programs of the control system are designed in detail, and the production of PCB and object is completed. The performance of the robot's obstacle avoidance is tested and verified, and the parameter basis is provided for the navigation strategy planning.
The basic principle of classical Calman filter, attenuated memory filter, limited memory filter and their combination usage are analyzed. On this basis, a limited memory Calman filter with exponential fading factor is proposed as a multi-sensor fusion method for coal mine disaster relief robot. The structure of the positioning system is introduced, and the autonomous positioning method based on the track estimation is introduced, and the improved method of track estimation is proposed. Using the limited memory Calman filter with exponential fading factor, the multi sensor information is fused by the progressive filtering method, and the minimization control of the error is realized. The STC12C5A6 is developed. The 0S2 micro controller is the core of the multi-sensor information fusion hardware system, and the software algorithm flow is given. The simulation results show that the filtering method can stabilize the tracking precision of the 0.5m. On this basis, the autonomous positioning function of the robot is tested, and the effectiveness of the method is verified.
Then, through the study of several matching methods, a quasi dense matching method based on optimal seed sorting is proposed in view of the real-time requirement of robot. The method of obtaining three-dimensional information of obstacles by single point method of pole line is discussed. First, the edge of obstacles is segmented, and the pole line is used to accelerate the matching, and the obstacle is obtained one by one. In order to find the camera coordinates of the maximum points of each edge, the 3D information of the obstacles - height, width and distance are obtained, and the mathematical model of the slope calculation is established. The RBF method is used to estimate the slope terrain. The Canny algorithm is used. The edge of continuous ladder image is detected, and the least square method is used to deal with the straight line.
Then, in order to improve the understanding ability of the coal mine detection robot to the navigation environment, according to the obstacle ability of the CUMT-5 type coal mine detection robot designed in this paper, and using the means of fuzzy theory, image processing technology and neural network technology, the analysis and evaluation method of the terrain availability are discussed, and the space position is put forward. Information and visual sensing information are fused from five aspects: terrain undulation, terrain opening, terrain slope, topographic discontinuity and continuous ladder influence. Based on the fusion of fuzzy theory, functional learning and neural network technology, the author creatively puts forward a new method. The behavior control navigation method based on the coordination strategy of arbitration is introduced, and the behavior controller is designed based on the typical environment model. The advantages of the fusion mechanism and the arbitration mechanism are integrated into the behavior coordination strategy, and the adaptive control of the basic behavior is realized. On this basis, the terrain can be evaluated. It is introduced into the coordination strategy to further realize the navigation of coal mine detection robot in complex and unstructured environment.
Finally, a comparative experiment on robot integrated navigation is carried out. The results show that the navigation method proposed in this chapter can realize autonomous navigation in complex and unstructured environment. It not only has better adaptability and reliability to different environment, but also has high energy constraint ability.

【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TD714;TP242

【参考文献】