【作者】 刘宇飞；

【导师】 刘光军； 高海波；

【作者基本信息】 哈尔滨工业大学 ， 航空宇航制造工程， 2019， 博士

【摘要】 多足步行机器人以运动稳定灵活、非连续支撑、落足点离散的特点,使其能够胜任常规的轮式和履带式机器人无法完成的工作,具有对非预知地形及崎岖地形的高度适应能力和运动灵活等特点。六足机器人作为足式机器人的一个典型代表,在极端工作条件下的物品运输、抢险救灾、极限作业等领域的应用需求日益明显。六足机器人在复杂地形行走过程中带来的问题包括:足-地作用过程中的运动规划算法多样、本体传感器框架下的六足机器人运动状态模糊、行走过程中的内力过大、足-地滑移影响运动效率、多传感数据融合复杂等,对六足机器人的传感融合及运动控制提出了更高的挑战。在该方向的研究为足式机器人的运动控制提供理论与技术支持。本文以电驱动六足机器人为研究对象,对带有被动弹性足的足端运动规划方法进行了研究。在设计过程中为了保证足-地之间的柔顺接触,在小腿和足端之间加入了被动弹簧结构,使六足机器人的足成为了可伸缩的弹性结构。分析了带有被动弹性足的机器人在行走过程中的足-地动态接触过程,针对传统规划方法带来的足-地法向及切向冲击问题,提出了基于被动弹性足地面作用的低冲击、低能耗足端运动规划方法,解决了带有被动弹性的足在抬起或落下过程中的切向和法向冲击及机器人本体频繁加减速带来的能耗增加问题,并在六足机器人样机上进行了验证。由于传感器本身的误差和安装误差带来的影响,以及足-地作用时的足端滑移或机械误差等,单独使用本体传感器无法得到准确的机器人速度、位姿等信息。针对此问题,基于扩展卡尔曼滤波技术对六足机器人的运动学和惯导信息进行融合,得到了机器人的本体速度和位姿等信息。当足式机器人发生足地滑移时,很难通过本体传感器检测到滑移的发生,基于单腿滑移的动力学模型,本文建立了螃蟹型行走及昆虫型行走的足地滑移加速度模型,提出了基于状态估计的足地滑移估计方法,检测了足地滑移率并对足端滑移状态进行了判定,通过与外部运动捕捉系统(Mocap system)得到的机器人状态数据进行对比,验证了所提出方法的有效性。为了实现六足机器人在非预知地形下的自主行走,对六足机器人在多腿支撑下的耦合特性进行分析,基于地形的估计得到力学转换矩阵并用于对足端受力方向进行修正。针对大摩擦地面,提出了稳态行走下的足力优化方法,解决了牵引和切向方向运动不同步而产生内力的问题。在非预知地形下,由于弹性足的变形、地面变形及内力对抗导致的足底滑移会导致机器人行走轨迹发生偏移并影响运动效率,支撑足的有效位移与预先规划的位移有明显的差距,提出了考虑地形估计及滑移的足力优化方法,以降低各足间的切向内力为目标,抑制由于内力对抗导致的足端滑移情况的发生,通过在非平坦及平地-斜坡过渡的非预知地形下验证了该方法的有效性。为了防止机器人在通过非预知地形时关节达到极限位置,本文建立了地形特征对六足机器人关节约束的数学模型,保证了机器人在不同地形下的关节运动空间安全性,根据地形几何约束模型为位姿控制提供俯仰和滚转目标角度。为了实现六足机器人的复杂地形的自主行走,基于外部视觉感知融合对六足机器人的自主避障及落足点选择策略进行了研究,建立了一种基于人工势场法的自主避障策略,通过视觉对地形高度、落足安全区域面积及地形倾斜滑移度进行分析并对落足点区域进行选择。对机器人的位姿控制、腿部控制进行了研究,在盲走和视觉引导状态下,采用了周期步态和连续跟踪步态,保证了机器人在复杂地形下的步序协调性。最后进行了六足机器人野外非预知地形的移动控制实验研究,验证了所提出方法的正确性和实用性。更多还原

【Abstract】 The multi-legged walking robot is characterized by its stable and flexible motion,non-continuous support and discrete foothold,making it capable of performing tasks that cannot be performed by conventional wheeled and tracked robots.It has high adaptability and flexibility to unknown and rugged terrain.As a typical representative of legged robots,hexapod robots have more demanding applications in the fields of transportation,disaster relief and extreme operations under extreme working conditions.The problems caused by the hexapod robots during the rugged terrain walking include: the flexible of motion planning algorithms in the process of foot terrain interaction;the state of the robot under the body sensor framework is not clear;large internal force and foot terrain slippage affect the walking efficiency;the motion control based on multi-sensor fusion is complex,which poses a higher challenge to the sensing fusion,locomotion control of the hexapod robot.The research in the subject promotes the locomotion control for the legged robots,which provides theoretical and technical support for the research of legged robot.The electrical driven hexapod robot is studied in this paper.The foot trajectory planning with passive elastic foot is studied.In order to ensure smooth contact for the foot terrain interaction,a passive spring structure is added between the leg and foot,so that the foot of hexapod robot become elastic structure with retractable.The foot terrain dynamic interaction process of a robot with deformable leg is analyzed.Aiming at the foot terrain impact force for normal and tangential directions caused by traditional planning methods,a low impact force and energy consumption foot trajectory based on passive elastic legs is presented.The method solves the foot terrain impact problem during the lifting and touching process by the passive spring.And the increase of the energy consumption caused by the frequent acceleration and deceleration of the robot is also solved.The method is verified on the hexapod robot prototype.The accurate of the velocity and posture data of hexapod robot can not be obtained by sensors only due to the influence of the error of the sensor itself and the installation error,as well as the foot slip or mechanical error during the foot terrain interaction.The data fusion method based on Extended Kalman Filter is adopted,and the data are from the kinematics and IMU information.The accurate velocity and posture of trunk body is obtained.When the slippage occurs,it is difficult to detect the occurrence of slip through the body sensors.This paper establishes foot terrain slip acceleration model for crab and insect walking based on leg dynamics model with foot slip,and presents foot terrain slip estimation method based on state estimation.The method can detect the slip rate of the foot and the state of the foot slip is determined.The estimated value is compared with robot state data obtained by the external motion capture system(Mocap system),and the effectiveness of the proposed method is verified.In order to realize the autonomous walking of the hexapod robot under the unknown terrain,the coupling characteristics of the multi-leg support are analyzed.The mechanical transformation matrix based on terrain estimation is obtained and is used to correct the foot force direction on the sloped terrains.The foot force optimization method for steady walking under large friction terrain is proposed,which solves the internal force generated by the unsynchronized locomotion in the traction and tangential directions.The foot slippage caused by the deformation of the elastic foot,deformation of unknown terrain and the large internal force will cause the offset of the walking trajectory and affect the locomotion efficiency.The effective displacement of the stance foot has significantly different from the pre-planned displacement.The foot force optimization method considering terrain estimation and foot slippage is proposed.The method reduces the internal force in the tangential direction,and the occurrence of foot slippage due to internal force confrontation is suppressed.The effectiveness of the method is verified by experimental data under complex and flat-slope transition unknown terrains.In order to prevent the joint from reaching the limit position under the unknown terrain,this paper presents mathematical model for the specific structural topography and joint space,which ensures the safety in the joint space.The pitch and roll target angles are adjusted according to the influence of the geometric constraint.The autonomous obstacle avoidance and the foothold selection strategy of the hexapod robot are studied based on external visual perception fusion.An autonomous obstacle avoidance strategy based on artificial potential field method is presented.The terrain height,safety landing area and the terrain inclination are analyzed and the area of the landing point is selected.The posture and leg control of the robot are studied.For the blind and visual guidance walking state,the periodic and continuous tracking gaits are adopted to realize the walking coordination under the complexed terrain.Finally,the experimental on unknown field terrains for the locomotion control of the hexapod robot are carried out,which verifies the correctness and practicability of the proposed method.更多还原