| 题名 | 连续体手术机器人感知与路径规划技术研究 |
| 作者 | 张如美 |
| 学位类别 | 博士 |
| 答辩日期 | 2017-11-30 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 沈阳 |
| 导师 | 赵忆文 ; 刘浩 |
| 关键词 | 连续体手术机器人，接触感知，形状感知，路径规划，精准操控 |
| 其他题名 | The Research on Sensing and Path Planning Technologies for Continuum Surgical Robots |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 本文在国家自然科学基金的资助下，从病人安全角度出发，针对连续体手术机器人的末端接触感知、本体形状精确跟踪、路径规划、基于多传感反馈的机器人控制等关键技术进行了深入研究，具体内容如下：研究了面向连续体微创手术机器人的末端接触感知技术。基于互感耦合感应原理，设计了四线圈式的三轴电磁形变传感器，利用接收与发射线圈之间互感的变化来对由外力引起的弹性体形变进行检测。针对接收线圈感应信号微弱这一特点，设计了相匹配的锁相放大电路用以提取微形变引起的感应电压变化量，并且探索了位移-互感-感应电压之间的关系，利用实验和仿真建立了其变化模型，给出了相应的形变解调算法。设计的电磁传感器可以用于检测连续体机器人末端执行器的形变变化及其与组织的接触关系，还可进一步与弹性体力学模型相结合用于手术器械末端力感知技术研究。针对FBG形状传感器定位时重建误差累积叠加，连续体手术机器人最常用的视觉感知方法易受光照、遮挡、阴影等环境影响鲁棒性差的问题，本文提出一种精确的、鲁棒的形状及末端位姿感知方法，利用扩展卡尔曼（EKF）融合算法对FBG形状数据和立体视觉数据进行融合来弥补两者各自的缺点。首先利用经验运动学模型求取机器人雅可比矩阵参数，开展实验对FBG形状定位及立体视觉定位的精度进行了评估，并将FBG定位误差、运动学模型误差与连续体长度的关系以分段的形式数值化呈现，使EKF可以按段对多传感器数据进行权重分配，进而提高连续体机器人本体跟踪精度。研究分别开展了两种实验用于评估系统的精确性及鲁棒性，实验结果表明基于EKF的融合能够提高系统的单一传感的跟踪精度，并且当视觉失效时，跟踪系统依然可以鲁棒的运行。研究了具有无限自由度的连续体机器人路径规划算法，提出了机器人运动过程中需要满足的运动学约束及可达性约束。由于材料特性，连续体机器人的运动轨迹需要满足曲率、挠率最大值限制及连续性约束，并且其轨迹规划必须保证连续体手术机器人末端执行器在目标点处的位姿指向靶点组织，以保证手术的成功。本文针对“follow-the-leader”模式的连续体机器人，基于反向快速扩展随机树 (BRRT) 算法提出了一种路径及轨迹规划算法。利用常曲率圆弧及Frenet框架下曲率与挠率几何定义对3D空间离散路径点的生长方向进行了约束，来满足其曲率挠率最大值要求。基于Frenet定理对离散点处的曲率挠率进行线性插值计算插值点处的切向量，对其积分完成运动曲线的轨迹重建。为了消除重建误差且使系统更易于控制，本文将3D空间规划降到2D平面内，利用排列组合算法依次去除多余的路径点，将Frenet公式与Bezier曲线相结合来获得平滑的且满足连续体机器人运动学和手术可达性限制的轨迹。最后本文搭建了连续体手术机器人的实验平台，提出了基于多传感反馈和逆运动学模型的机器人操控方法，利用立体视觉对手术场景进行了3D重建，建立了靶点组织及障碍区的立方体模型，并根据连续体机器人的结构尺寸设定合适的安全阈值，完成了连续体手术机器人末端与靶点的接触检测、本体形状与障碍区的碰撞检测实验。综上所述，本文从连续体手术机器人临床化面临的瓶颈问题出发，解决其难以精确操控所对应的感知方法匮乏限制，针对器械末端接触感知、本体形状精确鲁棒跟踪及手术中的路径规划等关键问题进行了深入研究，通过仿真和实验验证了所提出的设计或方法的可行性。这些研究为后续的连续体机器人人机交互、手术半自动/全自动、人机协作等技术的发展提供了理论基础。 |
| 英文摘要 | To overcome the limitations, this work is done under the national natural science foundation, focusing on the key technologies of continuum robots including distal touch perception, accurate shape tracking, path planning and robot control based on multi-sensor feedback. Details are as follows: This paper studied the distal touch perception technology for continuum robots. Based on the theory of mutual inductance coupling, a tri-axial electromagnetic deformation sensor with four-coil structure is designed. The deformation of elastomer caused by external forces is measured by using the mutual inductance variations between the four receiving coils and transmitting coil. The paper designed corresponding lock phase amplifier circuit to pick up the weak sensing voltage variations caused by micro-deformation. The relations between deformation, mutual inductance and sensing voltage were explored and the model was established, and the demodulation algorithms were derived based on simulation and experimental data analysis. The proposed electromagnetic sensor can measure the distal deformation information of surgical instrument and their interactions with human tissue. Furthermore, it can be combined with elastomer mechanical model to study the 3D force sensing technology of surgical instruments. For the continuum robotic sensing technologies, the reconstruction error of FBG shape senor is gradually increasing with the integral length, and the commonly used visual method has poor robustness and is easy to be distracted by illumination, occlusion, shadows, etc. To overcome these limitations, this paper presents an accurate and robust shape and distal pose tracking method for continuum robot by using extended Kalman filter (EKF) for fusion of FBG shape data and stereo vision data to compensate for their own shortcomings. Firstly, the parameters of robot Jacobian matrices are obtained by using the empirical kinematics model, and the positioning accuracies of FBG shape sensor and stereo vision were evaluated based on experiments. The relations between FBG positioning error, kinematical model error and continuum length were numerically presented in sections, so that the EKF can give them with different weights at different positions. This form can improve shape tracking accuracy.Two experimental cases are conducted to verify the system accuracy and robustness, and the results indicate that the fusion algorithm can improve the system tracking accuracy of a single senor and keep it working robustly when stereo vision is distracted during surgery. The paper proposed a path planning algorithm for continuum robots with infinite degrees of freedoms (DOFs), and explored the kinematic constraints and reachability constraints that need to be satisfied in the process of robot movement. Because of the characters of material, the trajectories followed by continuum robots must satisfy the maximum curvature and torsion constraints, and their continuity constraints. Besides, the end-effort should be pointed to the target tissue at the goal point to ensure the effectiveness of the surgery. This paper presents a path and trajectory planning algorithm based on back Rapidly-exploring Random Trees (BRRTs) for continuum robots working in a “follow-the-leader” manner. The maximum curvature and torsion restrictions were imposed by limiting the growth direction of random sampling points in 3D space based on piecewise constant curvature assumption and curvatures’ geometric definitions under Frenet coordinate. The trajectory was reconstructed by integrating the tangent vector at the discrete point which was calculated based on curvatures interpolation algorithm and Frenet formula. The paper reduced the 3D planning into 2D one to remove the error and reduce the control complexity. The extraneous waypoints were pruned based on permutation algorithms, and the trajectory was optimized by combining Frenet-based and third-order Bezier-based curve to achieve a smooth and kinematics and accessibility constrained path. Finally, an experimental platform for continuum robots was built. An accurate control method based on multi-sensor feedback and inverse kinematics was proposed. The 3D surgical scene was reconstructed based on stereo vision. The target tissue and obstacle area were chosen manually and assumed to be cuboid in 3D space. The safety threshold was set according to the geometric size of continuum robot. Both the contact detection experiment between distal point and target tissue and collision experiment between robot body and obstacle were carried out. In conclusion, this paper aims at the bottleneck issues of clinical application of continuum robots and focus on the problem of robot imprecise manipulation due to lack of perceptual methods, and mainly studied the key technologies of distal perception, accurate shape tracking method and path planning. The simulation and experiments were implemented to validate the effectiveness of the proposed designs or methods. Solutions to these limitations can provide a theoretical foundation for the subsequent associated techniques, such as human-robot interaction, semi-automatic or automatic surgery and human-computer collaboration. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 111页 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.sia.cn/handle/173321/21279]  |
| 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 | 张如美. 连续体手术机器人感知与路径规划技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所. 2017. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论