Traditional trajectory planning approaches are currently lacking in intelligence and autonomy. We used the reinforcement learning approach to solve the autonomous trajectory planning of the robot arm to avoid obstacles with uniform motion and hit the target point quickly with obstacle avoidance planning for surgical robots taken as the practical background. We used the algorithm of experience playback mechanism combined with off-policy DDPG based on reinforcement learning, and after several iterations, the robot completed trajectory planning with obstacle avoidance autonomously. Moving obstacles were added to roughly simulate the autonomous obstacle avoidance of a surgical robotic arm with moving medical personnel or mobile instruments in the operating room, based on the simple trajectory planning example of Open-AI Open-Source Project Baseline, combined with the research context. Sparse rewards were used for each iteration based on the HER algorithm, so that each attempt could gain experience. The HER-DDPG method can quickly complete the manipulator's trajectory planning in a simulation environment, which is critical for the surgical robot's autonomous positioning in the real world. Furthermore, the experience playback system has been tested to allow full use of sparse rewards and handle parallel tasks equally well.

Compared with traditional surgery, femtosecond laser minimally invasive surgery has many uncomparable advantages and will have a significant impact on the medical industry in the future. In this paper, a simple scene for laser minimally invasive virtual surgery training is designed, in which the testers can practice repeatedly until the basic operational requirements are met. The haptic device adopts the Geomagic Touch from American 3D Systems Company. Eight testers using Geomagic Touch handle perform four basic actions (clamping, adjusting posture, pushing / pressing, moving tiny objects) in the left interface virtual environment. Each action was performed 10 times by per tester. During the process of human-computer interaction, the position, attitude, speed, button and other information of the handle are collected in real time, and the collected data is saved in the form of text. The collected data is multivariate time series data. Based on the characteristics of multivariate time series data, this paper proposes a design of an evaluation system based on LSTM model to classify the collected data and evaluate the standard of surgical action according to the output probability of action classification.

提出一种基于遗传-拟牛顿(GA-BFGS)混合算法的Watt-Ⅰ型六杆机构优化算法,并针对多自由度下肢康复机器人结构复杂、控制困难、造价昂贵等问题,设计了两种单自由度Watt-I型六杆下肢康复机构.首先,由目标步态轨迹和Watt-I型六杆机构运动学模型构造两种目标函数;然后,结合遗传算法全局搜索性和拟牛顿算法局部细致寻优的特点,使用GA-BFGS混合算法对目标函数进行迭代寻优,得到优化后的机构尺寸;最后,分别将两种下肢康复机构应用于下肢康复机器人,并对其进行了结构设计.结果表明:两种下肢康复机构都能较好地复现目标步态轨迹,一种侧重于复现其时序,另一种侧重于复现其形状,这验证了该算法对于Watt-I型六杆机构轨迹综合问题的有效性.

The major drawbacks of current lower limb rehabilitation robots are high cost and complex structure which make them inappropriate to be applied in the community and family. In this paper, we design an 1-degree-of-freedom (DOF) robot with humanoid gait for lower limb rehabilitation based on Watt-I six-bar mechanism. Let the normal gait trajectory be target trajectory, the dimensions of the mechanism are calculated by path synthesis. First, the objective function to reflect the accuracy of trajectory reproduction and relevant constraints are established. Then GA-BFGS hybrid algorithm is used to minimize the objective function. After that, the optimized mechanism is analyzed by trajectory comparison, velocity / acceleration analysis and joint angle detection. Further, the kinematic simulation of the mechanism is also completed. The results show that while the crank is rotating at a constant speed, the mechanism can reproduce the time sequence and the shape of target trajectory approximately to realize walk training for patients with lower limb disorders whose legs are 810.0–860.0mm long (the corresponding heights are about 1650.0–1750.0mm). Finally, the specific structure of lower limb rehabilitation robot based on this mechanism is designed and the principle prototype model is given.

Computer-assisted cognitive training is an effective intervention for patients with mild cognitive impairment (MCI), which can avoid the disadvantages of traditional cognitive training that consumes a lot of medical resources and is difficult to be standardized. However, many computer-assisted cognitive training systems have unfriendly human-computer interaction, for not considering that most MCI patients have certain difficulties in using computers. In this paper, we design a cognitive training system which allows patients to implement human-computer interaction through gestures. First, a gesture recognition algorithm is proposed, in which we implement gesture segmentation based on YCbCr color space and Otsu algorithm, extract Fourier Descriptors of gesture contour as feature vectors and use SVM algorithm to train a classifier to recognize gestures. Then, the graphical user interface (GUI) of the system is designed to realize the task requirement of cognitive training for the MCI patients. Finally, the results of tests show the accuracy of the algorithm and the feasibility of the GUI. With the above computer-assisted cognitive training system, patients can achieve human-computer interaction only through gestures without the need to use keyboard, mouse, etc., greatly reducing the burden of patients during training.