中国科学院自动化研究所机构知识库

随着机器人技术的发展，目前在工业领域中，传统机械臂已经能够代替人工执行操作简单、重复性强的作业任务。但仍然存在一些复杂狭窄的场景，传统工业机器人很难开展作业，例如核设施检测、飞机零部件装配、电力设备维护等。同时这些场景中，人工作业的方式，要么很难保证作业质量和作业安全，要么因为环境中存在高温、高辐射、有毒有害气体等情况，根本无法进行现场作业。这些场景的共同特点是作业空间复杂狭窄、障碍物分布杂乱、工作环境恶劣、作业点位于空间深处等。因此，针对复杂狭窄空间中作业型机器人的研究具有很重要的理论意义与应用价值。本文以具有大长径比、超冗余自由度的机械臂为对象，在机构设计与实现、运动学建模、张力建模与优化以及避障路径规划等方面开展研究，主要内容如下：

第一，针对复杂狭窄空间中的作业需求，研制了一种基于多自由度关节的超冗余绳索驱动蛇形机械臂。利用关节的多自由度设计，提高了机械臂的灵活性。通过模块化的关节和连杆结构，达到了不同作业场景下机械臂的灵活配置。针对绳索驱动器高精度、高负载、小体积的要求，研制了基于梯形丝杠的直线式绳索驱动器。针对狭窄空间的作业需求，优化了驱动器布局，将它们按筒状结构分内外两圈排列。最后，搭建了蛇形机械臂控制系统，形成了10关节蛇形机械臂实验平台。

第二，针对绳索驱动超冗余机械臂的运动控制问题，提出了一种由驱动空间到位姿空间的运动学建模方法。通过分析机械臂位姿、关节角、绳索长度以及电机驱动位置之间的关系，建立了运动学模型。针对工作空间求解时运动学逆解困难的情况，利用正运动学建立了基于蒙特卡罗方法的工作空间求解模型。在此基础上，通过分析连杆长度、关节最大转向角等机构参数对工作空间的影响，对超冗余机械臂的机构参数进行了优化。

第三，针对绳索驱动下的张力估计与优化问题，提出了一种绳索张力建模与优化方法。基于刚体力矩平衡原理，建立了绳索张力模型，并按照关节机械约束进行力矩分解，实现了绳索张力的求解。利用机械臂的冗余自由度，在路径规划中根据张力模型构建张力优化评价函数，实现了绳索张力的优化。

第四，针对超冗余机械臂的避障路径规划问题，提出了一种改进人工势场法与关节跟随避障相结合的路径规划方法。通过将路径规划过程分解为两个阶段，降低了规划的复杂度。在第一阶段，针对机械臂的末端引导轨迹搜索，提出了一种基于被动斥力机制的改进人工势场法，避免了引力函数和斥力函数失配导致的问题。在第二阶段，针对机械臂末端沿引导轨迹前进时其余关节位姿求解的问题，通过关节和连杆之间的几何约束关系构建了适应度函数，提出了一种基于最优搜索的关节跟随方法。同时，为实现避障功能，提出了一种基于夹角的碰撞检测方法，提高了碰撞检测效率。

第五，针对绳索驱动的蛇形机械臂几何参数标定问题，提出一种基于关节旋转角的参数辨识方法。利用对旋转角的测量代替对关节末端位姿的测量，克服了测量过程复杂、设备昂贵的难题。以关节角为中间变量，分析了几何参数与旋转角之间的关系，建立了几何参数辨识模型，实现了关节几何参数标定。针对旋转角测量问题，将单目相机和加速度传感器的信息通过扩展卡尔曼滤波进行融合，提出了一种操作简便、成本低廉的旋转角测量方法。基于上述辨识方法和测量方法的标定过程，具有较高的操作便利性和环境适应性。

最后，对本文的工作进行了总结，并指出了需要进一步开展的工作。

With the development of robotics, currently, in the industrial field, traditional manipulators have been able to replace humans to perform simple and repetitive tasks. However, there are still some operation scenarios with complex and narrow space that is difficult for traditional industrial robots to work within it, such as nuclear facility inspection, aircraft assembly, and power equipment maintenance. At the same time, in these cases, for manual work, either it is difficult to guarantee the quality and safety or because of the high temperature, high radiation, and toxic and harmful gases in the environment, it is impossible to conduct on-site operations at all. The common characteristics of these scenes are the complex and narrow working space, the disorderly distribution of obstacles, the harsh working environment, and the working point is located deep in the space. Therefore, the study of robots suitable for complex and narrow spaces has very important theoretical significance and application value. This paper takes the manipulator with a large aspect ratio and hyper redundant degrees of freedom as the object and conducts research in the aspects of mechanism design and realization, kinematics modeling, tension modeling and optimization, and obstacle avoidance path planning. The main contents are as follows:

Firstly, in response to the work requirements in complex and narrow spaces, a hyper-redundant cable-driven snake-like manipulator based on multi-degree-of-freedom joints was developed. The multi-degree-of-freedom design of the joints improves the flexibility of the manipulator. Through the modular joint and link structure, the flexible configuration of the manipulator under different operating scenarios is achieved. For the requirements of high precision, high load, and small size of the cable driver, a linear cable driver based on a trapezoidal screw has been developed. Due to the narrow space, the cable drivers' layout is optimized, and they are arranged in a cylindrical structure in two circles inside and outside. Finally, a snake-like manipulator control system was built to form a 10-joint snake-like manipulator experimental platform.

Secondly, aiming at the motion control of cable-driven hyper-redundant manipulators, a kinematics modeling method from driving space to pose space is proposed. The kinematics model was established by analyzing the relationship between the manipulator's posture, joint angle, cable length, and motor position. Due to the difficulty in solving inverse kinematics, a workspace solution model based on the Monte Carlo method is established by using forward kinematics. On this basis, the mechanical parameters of the manipulator were optimized by analyzing the influence of the link length, the maximum steering angle of the joint, and the number of joints on the working space.

Thirdly, aiming at the problem of tension estimation and optimization under cable drive, a cable tension modeling and optimization method is proposed. Based on the principle of rigid body torque balance, a cable tension model is established, and the torque is decomposed according to the mechanical constraints of the joints to solve the cable tension. A tension optimization evaluation function is constructed according to the tension model in the path planning and by the redundant degrees of freedom of the robotic arm,  the optimization of the rope tension is realized.

Fourthly, aiming at the obstacle avoidance path planning problem of the hyper-redundant manipulator, a path planning method based on the fusion of improved artificial potential field method and joint following obstacle avoidance is proposed. By decomposing the path planning process into two stages, the complexity of planning is reduced. In the first stage, aiming at the search for the end-guided trajectory of the manipulator, an improved artificial potential field method based on the passive repulsion mechanism is proposed, which avoids the problem caused by the mismatch of the gravity function and the repulsion function. In the second stage, in order to plan the other links' poses when the end of the manipulator is advancing along the guiding trajectory, a joint following method based on the optimal search is proposed and the fitness function is constructed through the geometric constraint between the joint and the link. At the same time, in order to avoid the obstacle, a collision detection method based on the angle between obstacles and links is proposed, which improves the efficiency of collision detection.

Fifthly, aiming at the geometric parameter calibration of the cable-driven snake-like manipulator, a parameter identification method based on the rotation angle of the joint is proposed. The use of the measurement of the rotation angle instead of the measurement of the posture of the joint end overcomes the problem of the complex measurement processes and expensive equipment. Taking the joint angle as the intermediate variable, the relationship between the geometric parameters and the rotation angle is analyzed and the identification model is established, and the joint geometric parameter calibration is realized. For the rotation angle measurement, the information of the monocular camera and the acceleration sensor is fused through the extended Kalman filter, and an easy and low-cost method of rotation angle measurement is proposed. The calibration process based on the above identification method and measurement method has high operating convenience and environmental adaptability.

Finally, the research work of this paper is summarized and pointed out the work that needs to be carried out in the future.