【作者】 崔馨丹；

【导师】 蔡鹤皋； 赵杰；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2013， 博士

【摘要】 模块化机器人是由大量的具有一定运动和感知功能的基本模块相互连接而成，能够构成多种构型完成多种任务的复杂机器人系统。与传统机器人相比，模块化机器人具有模块化、自变形和自修复等特点，同时，由多个自带关节的模块组合在一起的机器人系统具有一定的冗余度，灵活性强，使其具备了较强的运动能力，可以通过运动构型及运动模式的改变来适应不同的路面条件。然而，模块化机器人运动的超冗余度、运动构型多样性以及运动多模式化的特点，也使运动规划变得更加复杂。因此，寻找一种能够将模块化机器人多种运动构型的多种模式运动统一规划，使运动模式能够随着机器人构型改变而自适应变化，并且能够对外部环境反馈作出相应步态调整，以实现非结构化环境中的自主运动的控制方法对模块化机器人的发展具有重要的研究意义。模块作为机器人的基本单元，它的结构特点对整体构型的运动方式有着较大影响。本文采用正方体结构的UBot模块作为研究对象，模块分别主动模块和被动模块，分别带有主、被动钩爪式连接机构，主动模块与被动模块可以实现自动的连接/断开，通过连接机构上的方位识别单元识别模块间的连接状态及方位，具有阵列式机器人模块结构规整、连接机构灵活的特征；每个模块具有两个相互垂直的旋转自由度，旋转角度范围为(-90°，90°)，可以实现串联式机器人整体协调运动的功能。为了使机器人系统具有环境感知功能，为UBot系统设计了传感模块，传感模块内装有摄像头、红外传感器、线性霍尔传感器及加速度计等传感装置，并具有与UBot被动模块相同的连接机构，可以连接到运动构型中实时感知外部环境。借鉴生物运动神经系统的控制机制，建立模块化机器人仿生网络控制结构。针对UBot模块双关节间相互约束的结构特点，建立了能够控制模块两个关节同时运动的双输出中枢模式发生器（CPG）振荡模型，构建了双输出CPG调相运动控制网络，用于模拟生物的脊髓神经控制系统，并引入中间神经元、感知神经元和运动神经元，形成模块化机器人仿生运动控制网络。基于建立的仿生运动控制网络，分别针对模块化机器人所构成的两类构形——无肢体运动构型和有肢体运动构型制定运动规划策略。通过模拟低等生物将低级神经中枢的自激行为直接作用到各个关节的运动控制方式，对模块化机器人几种典型的无肢体运动构型进行运动规划。利用CPG调相控制器中节点连接方式对相位输出结果的影响，在网络节点数目不变的情况下，实现了蛇形构型蠕动、多节虫运动、蜿蜒运动、滚动和转弯运动的多种运动模式的统一规划；并解决了十字构型的直行、转弯运动模式，环形构型的滚动运动模式的运动规划问题；通过模拟高等动物运动神经控制系统的方法，针对有肢体运动构型提出了利用CPG调相控制器模拟高等动物的脊髓神经系统，为各个关节分别建立运动神经元的运动规划策略，面向由髋膝踝关节组成的有肢体运动构型制订了统一的运动神经元设计办法，实现了仿生四足构型对角步态和慢走步态的统一规划。从拓扑构型改变与拓扑构型不变两个方面，研究了模块化机器人构型改变带来的运动规划方法协调变化问题。对于由模块连接方式改变而产生的拓扑结构变形，考虑单链式构型的拓扑改变与运动规划，归纳出开链式结构与闭链式结构之间的变形条件和各自的运动特点；通过调整仿生控制网络中的CPG节点连接方式，建立能够同时适应开链式构型蠕动运动及闭链式构型滚动运动的规划方法。通过环形构型与蛇形构型变形与运动模式协调改变的仿真，验证了方法的有效性。对于模块化机器人拓扑结构不变，而运动功能改变的变形，考虑有肢体运动构型及其衍生的无肢体蠕动构型的运动功能改变与协调运动规划，在仿生运动控制网络框架不变的情况下，制订了网络节点参数的协调改变策略。通过仿生四足构型与其衍生的横向蠕动H型、向前蠕动H型之间的变形与运动模式协调改变的仿真，表明有肢体运动构型及其衍生的蠕动构型运动功能改变与运动协调规划方法的可行性。为了实现机器人在非结构环境下的运动功能，模拟生物运动神经系统调控功能的三级等级结构，建立了一套包含信息采集单元、底层局部信息反射规划器以及高层信息处理系统并具有环境反馈感知功能的运动控制器，使机器人根据外部环境的反馈，能够自主地完成越障、避障、上下坡及步态变换，并能够针对特殊环形实现自主变形以适应外部环境的需要。建立模块化机器人运动与环境反馈实验系统，进行蛇形构型、十字构型、环形构型及仿生四足构型多种模式的协调运动实验，验证仿生运动控制网络对多种运动模式统一控制的能力；进行环形构型滚动模式到蛇形构型蠕动模式的转化实验，验证运动控制体系对拓扑构型改变的协调适应能力；在设置了上下坡、路面凸起、物块阻挡及拱形门等障碍的环境中，进行机器人自主应对环境的运动试验，验证了机器人对环境的感知及自主适应能力。更多还原

【Abstract】 Modular robot is a complex robot systemcomposed of a large number of basic modules with motor and sensory function, which can transform into varieties of configurations to complete multiple tasks. Through the change of connected relation between modules, it can transform into varieties oftopological configurationto adapt to the environment and task requirements of various situation. Compared with the traditional robot, modular robot has features of modularity, self-reconfiguration and self-repair, etc.Meanwhile the modular robot system composed ofpluralities of modules has the characteristics of redundancy and flexibility,therefore, the modular robot has strong movement ability adapting to different road condition through movement configuration and locomotion mode change. However, hyper-redundancy, motion configuration diversity and multi-mode movement also make motion control of the modular robot much more complex. It is the research priorities of this paper to find a way of controlling multi-mode movement of the movement configuration of modular robot and making the corresponding gait adjustment to external feedback. Furthermore, this method should make the modular robot be able to exercise in unstructured environments.As the basic unit of the robot, structure characteristic of modules has great influence on motion mode of overall configuration. This paper takes the UBot module of cube structure as study object. It can be divided into active module and passive module, respectively, with the active and passive hook-claw type bindiny mechanism, the active module and passive module can achieve connection/disconnection automatically. Modules could identify direction and connection state, though orientation recognition unit on bindiny mechanism, which has the regular module structure and flexiblebindiny mechanismcharacteristics of array-type robots. Each module has two mutually perpendicular rotational freedom, the rotation direction (-90°,90°), so it can achieve the function of coordinated movement of chain type modular robots. In order to make the robot system capable of environment perception, designing the sensing module for UBot system, which is equipped with a camera, infrared sensor, linear Holzer sensor and accelerometer sensors, etc. Meanwhile, it has the same bindiny mechanism with the passive module of UBot, and could be connected with the active module to join onto movement configuration in real-time sensing of the external environment.Referencing the biological motion controlnervous system, it establishes motion control system of Bionic network, which consists of various kinds of neurons and the central pattern generator (CPG) composed by neurons. According to the characteristics of the UBot module, it also establishes a dual output CPG phase motion control network, to simulate neural control system of animal’s spinal cord and build up robot’s motion control system of Bionic network by introducing the intermediate neurons, efferent neurons, and Perception neurons, etc.Based on the establishedmotion control system of Bionic network, motion planning strategies are established for configurations with limbs and configurations with no limbs, respectively. Through simulating the nerve center of lower organisms’ movement control method sending the lower self-excited behavior directly to each joint, motion planning has been done for several typical configurations with no limbs. For CPG phase modulation control network, different CPG node connection can lead different phase timing output, using this characteristic, we have planned the uniform CPG node connection method for snake configuration to achieve mantis movement, wriggling, rolling, turning and multi-segmented worm-like locomotion without changing the quantity of the CPG nodes. We have also successfully achievedforward motion and turning motion plan of cross configuration, rolling motion plan of loop configuration through this bionic network. Using CPG phase modulation control network to simulate spinal cord nervous system of higher animal and making use of Efferent neurons to each joints, motion strategy has been proposed for configurations with limbs. For the configuration with limbs which has hip, knee and ankle joints on each limb, efferent neuron’s model has been designed for each joint. The unified planning for trot and walk motion of bionic quadrupedconfiguration is implemented.From topology changed and topology unchanged aspects, motion planning coordinating changing methods with configuration deformation have been studied. For the topology changed deformation, single-chain configuration’s topology changing method and motion plan has been researched, and deformation conditions from open-chain structure to closed-chain structure have been obtained. Through adjusting CPG nodes’ connection of bionic network, the motion planning method which can adjust open-chain squirming and closed-chain rolling has been established. Through the simulation of loop and snake configuration’s deformation and motion, the method has been proved. For the topology unchanged, but motion function changed deformation, the configuration with limbs has been studied. Bionic motion network’s node parameter coordinatedchanging strategy has been formulated for the deformation from the configuration with limbs to its derivative squirming configuration. Adaptive coordinated motion simulation between bionic quadruped configuration and its two derivativeH configurations has been conducted to prove the method.In order to realize the motion function of the robot in unstructured environment, simulate the three-level structure of regulatory function of biological neural system, it establishes a set of motion controller with the sensing function of environmental feedback, composed of information acquisition unit, underlying local information reflecting planner, and High-level information processing system, which makes the robot autonomously accomplish reconstruction of special situation and finish the motion of obstacle crossing, obstacle avoidance, self-reconfiguration, up and down the slope, so as to adapt to the external environment.Experiment system of movement and environmental feedback is constructed, and coordinated motionexperiments of multiple modes of snake-like configuration, cross configuration, loop configuration and bionic quadruped configuration are carried out, to test and verify overall control ability of the control system of Bionic network in different motion mode. The transformation experiment from rolling mode of loop configuration to peristalsismode of snake-like configuration are conducted, to test and verify coordination and adaptation ability of bionic motion control system for the topological configuration transformation. Motion experiment of autonomic robot response to environment in the environment with obstacle of slope, road bump, block and arch, etc, are finished to test and verify environmental perception and adaptation ability of the modular robot.更多还原