【作者】 徐兆红；

【导师】 吕恬生；

【作者基本信息】 上海交通大学 ， 机械电子工程， 2009， 博士

【摘要】 腿式跳跃机器人具有很强的移动性、灵活性和地形适应性,是移动机器人研究领域中的一个前沿应用课题,在家庭服务、医疗娱乐、星际探测、军事侦察、反恐救灾等领域有着极其广泛的应用前景。腿式跳跃机器人是一个非线性、多变量、强耦合和变结构的复杂动力学系统,其驱动性能约束下的逾障规划和动态平衡问题的研究具有很大的挑战性。本文以不带弹性元件、完全电机驱动的多关节腿式跳跃机器人为研究对象,从提高机器人运动能力的角度,结合驱动约束条件,对腿式跳跃机器人的动力学建模、运动规划与优化、稳定性恢复控制等关键问题进行了研究,并通过仿真实验和机器人实体试验进行了验证。论文的主要工作如下:(1)腿式机器人跳跃运动的特征量提取基于生物的跳跃形态和人体跳跃运动捕捉试验,从生物力学、仿生学和机器人学的角度分析了影响跳跃运动的主要因素,采用起跳姿态、广义负载和上肢运动三个特征量刻画机器人的跳跃运动性能。(2)腿式跳跃机器人的动力学建模跳跃运动的腾空相参考基座是浮动基,具有动量矩守恒的非完整约束;站立相的参考基座是固定基,具有完整约束。针对变约束特性,采用Lagrange浮动基方法,推导了跳跃机器人的变约束动力学模型。(3)腿式机器人的跳跃逾障运动规划为提高机器人的地形适应能力,给出了基于模型综合与解耦策略的腿式机器人跳跃逾障运动规划方法。该方法先将腿式机器人的上身等效为一个具有惯量特性的刚体,根据跳跃姿态和障碍物,运用内在动力学,规划运动综合模型的跳跃逾障运动;接着基于等效刚体的COM可操作度,运用开环可操作度优化方法进行冗余度分解,规划冗余关节的运动。(4)腿式机器人跳跃运动的特征量优化视腿式跳跃机器人为持有末端载荷的冗余机械臂,从力与运动传递性能的角度,结合起跳动力学约束条件,给出了跳跃运动的特征量优化方法。该方法运用惯性匹配方向可操作度优化了跳跃运动的三个特征量,通过特征量的优化提高了跳跃性能。惯性匹配方向可操作度作为一种动力学评价指标,度量了机器人在跳跃任务下的跳跃高度。(5)腿式跳跃机器人的稳定性恢复控制基于驱动性能约束,给出了落地冲击力作用下的稳定性恢复控制方法。该方法运用ZMP可操作椭圆规划了跳跃步态,当机器人有足够的驱动能力进行自运动姿态调整时,采用ZMP平面映射法进行稳定性恢复;当机器人的驱动能力不足时,通过步态调整使机器人重获稳定。(6)人体运动捕捉与机器人跳跃运动试验为了验证本文所给出方法的有效性,分别构建了人体运动捕捉和机器人跳跃运动的试验平台。利用三维图像检测系统,通过人体跳跃运动的捕捉,分析了影响跳跃性能的主要特征因素;研制了两类跳跃机器人试验样机,分别验证了跳跃运动特征量优化方法的有效性,以及落地冲击下稳定性恢复控制策略的实用性。本文有关跳跃机器人的动力学建模、运动规划、运动优化以及地面冲击力作用下的稳定性恢复控制方法,有助于提高机器人的灵活性和运动能力,拓展机器人的应用领域,在理论和应用上都具有一定的借鉴作用和参考价值。更多还原

【Abstract】 In complex or nonstructure environment, legged jumping robots have a better mobility, dexterity and terrain suitability than wheeled or tracked robots. As an advanced topic in mobile robots, they have well application foreground in fields of house work, rehabilitation treatment, sports science, life entertainment, interstellar exploration and national defence.Legged jumping robot is a multivariable, strongly coupled, nonlinear and varying structure dynamics system. Due to actuator constraints limitation, the ability of jumping over obstacles and dynamic stability are very challenging for control theory and motion planning. This dissertation concentrates on multi-joints legged jumping robot without elastic actuators. In order to improve versatility of jumping motion, some key problems related to dynamics modeling, motion planning, motion optimization and stability recovery are studied. Simulations and experiments have also been performed to test the methods presented in the dissertation. The discussed contents are listed as follows.(1) Character parameters of jumping motionFrom biology forms under literature datum and human jumping motion under motion capture experiment, three character parameters can be used to describe jumping robots related to biomechanics and robotics, and they are take-off postures, generalized load and upper limbs swing.(2) Dynamics modeling of legged jumping robotLegged jumping robot belongs to a variable constraint system because every phase has different constraint conditions. Stance phase has holonomic constraint, nevertheless flight phase has holonomic and non-holonomic constraints. The unified dynamics modelings of legged jumping robot including flight phase, stance phase and landing impact phase are achieved using Lagrange method based on floating basis.(3) Motion planning of jumping over obstaclesUsing the idea of motion synthesis and redundant DOFs decoupling, a motion planning method related jumping over obstacles is presented. Upper body is equivalent to a rigid body with inertia property. Based on internal dynamics and boundary conditions including jumping postures and obstacle’s size, a motion of jumping over obstacles is planned. Moreover, redundant DOFs are decoupled using open-manipulability optimization method based on COM manipulability of upper limb.(4) Motion optimization for legged jumping robotLegged jumping robot can be regarded as a redundant manipulator with a load at the end-effector. From the take-off dynamics and transfer performance between motion and force, inertia matching directional manipulability is used to optimize character parameters of jumping motion including take-off posture, generalized load and upper limbs swing. Inertia matching ellipsoid is an index of dynamic performance, and it measures the jumping height under jumping tasks.(5) Stability recovery under landing impactDuring landing impact phase, there is a large impact force. Aim at this landing impact, ZMP manipulability is introduced to plan jumping gaits based on actuator constraints limitation. Stability can be recovered by ZMP plane projection method within actuator performance, and it is only recovered through gait adjustment beyond actuator performance.(6) Experiments of human motion capture and robot jumping motionIn order to verify the proposed methods, human motion capture and experiment platforms for jumping motion are constructed respectively. Main factors are obtained by analyzed motion of key points based on human motion capture system. Motion optimization and stability control strategy are tested under two jumping robots which one is run linked to a 3D platform and the other is an autonomous robot.The main contributions of this dissertation consist in the improvement of mobility and jumping performance, motion planning and stability recovery control of legged robot. The presented methods are worth to be used for widening application fields and its reference value on the theory and application.更多还原