基于HCCPG模型的蛇形机器人控制方法研究

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 机器人学研究室

题名	基于HCCPG模型的蛇形机器人控制方法研究
作者	杨贵志
学位类别	博士
答辩日期	2014-11-29
授予单位	中国科学院沈阳自动化研究所
授予地点	中国科学院沈阳自动化研究所
导师	马书根 ; 王越超
关键词	蛇形机器人层次化联结CPG模型三维步态自适应逻辑神经元
其他题名	HCCPG-based Control of a Snake-like Robot
学位专业	机械电子工程
中文摘要	蛇形机器人具有多个运动自由度、运动灵活，从而具有能很好地适应环境的潜力和广泛的应用前景，比如地震废墟搜索救援等。但是目前的蛇形机器人的环境自适应能力还比较差，无法投入实际应用。因此需要继续深入研究以提高机器人的环境适应能力。蛇具有很强的环境适应能力的一个重要原因是它们能够利用身体的灵活性实现灵活多样的步态，并且能根据步态的特性以及所处的环境状态及时调整步态。因此要提高蛇形机器人的环境适应能力需要解决步态生成、步态分析、步态运用这3方面的问题。因此，本文的研究内容主要包括以下3个方面： (1) 提出HCCPG模型以控制机器人生成步态。蛇形机器人的步态基本都是节律的，而神经生理学的研究表明在脊椎动物中存在一种称为中枢模式生成器(Central Pattern Generator，CPG)的神经机制控制着很多节律性运动。因此可以通过构建CPG模型来提高机器人控制能力。因为CPG是由大量神经元组成的分布式系统，所以可以用神经元模型来构建联结CPG模型(Connectionist CPG，CCPG)以控制机器人。CCPG具有适合分布式实现、便于引入传感器感知信息等优点。因此，CCPG适合作为机器人的控制器。但是传统的CCPG模型在生成对适应环境具有重要意义的三维步态时，还有一些重要不足。本文根据CPG生物学原理，提出了一种运动神经元模型和一个构造层次化模型的方法，进而提出一个层次化CCPG模型(Hierarchical CCPG，HCCPG)。HCCPG模型解决了传统CCPG模型的步态生成问题。为更好地利用提出的模型，本文还对神经元模型、HCCPG模型的特性进行了深入分析。在HCCPG模型基础上，本文提出了一个二维步态、三维步态统一的生成方法。仿真验证表明该方法有效。 (2) 对三维步态的特性进行深入分析。因为三维步态是机器人适应复杂环境的基础，而且每种步态都有其适应的环境范围，所以需要对各种步态的特性进行深入分析，以提高它们的环境适应能力。由于机器人具有多个运动自由度，所以可以实现多种步态，但是无法一一进行分析。本文分析了3种主要的三维步态，即转弯步态、侧向蜿蜒步态、侧向翻滚步态。关注的步态特性包括运动特性和构型特性(身体外形特性)。运动特性即为步态参数、环境等对运动性能(速度、功耗等)的影响，而构型特性是指步态参数对机器人构型参数的影响。为了分析构型特性，本文提出了一种基于多连杆模型的分析方法。相比于传统方法，这种方法具有便于得到步态参数与构型参数之间的直接关系的优势。得到的步态特性为设计自适应控制器提供了依据。 (3) 提出基于HCCPG模型与侧向翻滚步态的环境自适应控制方法。坡面是个常见环境，但是现有的自适应方法不能很好地解决该环境下的适应问题。通过分析比较，本文选择以侧向翻滚步态为基础来适应该环境。为此，首先分析了机器人在坡面上翻滚时的特性，并将这些特性与HCCPG模型相结合提出自适应控制器。仿真、实验表明该方法比传统方法更有效。此外，还提出一个方向控制方法和一个构造逻辑控制系统的方法解决侧向翻滚时的路径跟踪问题。综上所述，针对蛇形机器人自适应控制中存在的问题，分别从HCCPG模型的建模与分析、统一步态生成方法、步态特性分析、坡面自适应控制方法等方面进行研究，从而提高蛇形机器人的环境适应能力。
索取号	TP242/Y27/2014
英文摘要	The snake-like robot has many DOFs, and thus has possibility to adapt agilely to environmental changes. Therefore, it has potential applications like searching and rescuing in the earthquake ruins. However, because of no sufficient method for the robot to achieve this adaptability, it is far from real applications. The robot's adaptability could be improved from three main aspects, namely the gait generation, the gait analysis and the gait utilization. Consequently, this dissertation includes three main contents. (1) An HCCPG model is proposed to generate the gaits for the snake-like robot to adapt well to the environment. Since the CPG is a distributed system composed of lots of neurons, the neuron models are used to compose the Connectionist CPG model (CCPG) to control the robots. CCPG has merit such as it is suitable for implementing the distributed control and is also convenient for integrating the sensory information. Therefore, CCPG is suitable for controlling the robots. The traditional CCPGs, however, have problems such that they can not generate the 3D gaits well that are important for the adaptation. In the thesis, we propose a hierarchical CCPG model (HCCPG) based on a motoneuron model and a layered CPG model. HCCPG model solves the gait generation problems in the traditional CCPGs. To make better use of HCCPG, the properties of the motoneuron and HCCPG are thoroughly investigated. A unified method for generating the 2D and 3D gaits are proposed based on HCCPG. The validity of the method is verified by the simulations. (2) The properties of the 3D gaits for snake-like robot are analyzed. The 3D gaits are the foundation for adaptation in the complex environment. Each gait can only adapt to its corresponding environment. Because the robot has many DOFs, it can implement many types of gaits. However, it is impossible to discuss all of them. We thus only analyze three main gaits, namely the turning, the sidewinding and the lateral rolling. Two kinds of properties, including the locomotion property and the configuration property, are concerned here. Locomotion property is a dynamic property and shows how the gait parameters and the environment affect the locomotion performance like the speed and the power consumption. On the other hand, the configuration property is a static property and shows how the gait parameters affect the configuration. To solve the problems in the traditional configuration analyzing method, we propose an analyzing method based on the robot's kinematics model. The proposed method can derive the direct relationship between the gait parameters and the configuration. The obtained gait properties provide useful instructions for designing adaptive controllers. (3) An adaptive control method is proposed based on the lateral rolling gait. The slope is a common environment. However, the existing adaptation method can not solve the adaptation problem well. The rolling gait is considered because of its excellence of motion in this environment. By analyzing the rolling motion on the slope, we propose an adaptation rule and design an adaptive controller by integrating the rule into the HCCPG model. Simulations and experiments validate this method. Moreover, we also propose a turning method and a method to construct a logical control system, to solve the path following problem for the rolling gait. In short, to improve the robot's adaptability, we carry out researches on several aspects, such as constructing and analyzing HCCPG model, proposing uniform gait generation method, analyzing gait properties and proposing slope adaptation method.
语种	中文
产权排序	1
页码	104页
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/16792]
专题	沈阳自动化研究所_机器人学研究室
推荐引用方式 GB/T 7714	杨贵志. 基于HCCPG模型的蛇形机器人控制方法研究[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2014.