BACKGROUND:Hemiplegia,a prevalent stroke-related condition,is often studied for motor dysfunction;however,spasticity remains under-researched.Abnormal muscle tone significantly hinders hemiplegic patients’walking rec...BACKGROUND:Hemiplegia,a prevalent stroke-related condition,is often studied for motor dysfunction;however,spasticity remains under-researched.Abnormal muscle tone significantly hinders hemiplegic patients’walking recovery.OBJECTIVE:To determine whether early suspension-protected training with a personal assistant machine for stroke patients enhances walking ability and prevents muscle spasms.METHODS:Thirty-two early-stage stroke patients from Shenzhen University General Hospital and the China Rehabilitation Research Center were randomly assigned to the experimental group(n=16)and the control group(n=16).Both groups underwent 4 weeks of gait training under the suspension protection system for 30 minutes daily,5 days a week.The experimental group used the personal assistant machine during training.Three-dimensional gait analysis(using the Cortex motion capture system),Brunnstrom staging,Fugl-Meyer Assessment for lower limb motor function,Fugl-Meyer balance function,and the modified Ashworth Scale were evaluated within 1 week before the intervention and after 4 weeks of intervention.RESULTS AND CONCLUSION:After the 4-week intervention,all outcome measures showed significant changes in each group.The experimental group had a small but significant increase in the modified Ashworth Scale score(P<0.05,d=|0.15|),while the control group had a large significant increase(P<0.05,d=|1.48|).The experimental group demonstrated greater improvements in walking speed(16.5 to 38.44 cm/s,P<0.05,d=|4.01|),step frequency(46.44 to 64.94 steps/min,P<0.05,d=|2.32|),stride length(15.50 to 29.81 cm,P<0.05,d=|3.44|),and peak hip and knee flexion(d=|1.82|to|2.17|).After treatment,the experimental group showed significantly greater improvements than the control group in walking speed(38.44 vs.26.63 cm/s,P<0.05,d=|2.75|),stride length,peak hip and knee flexion(d=|1.31|to|1.45|),step frequency(64.94 vs.59.38 steps/min,P<0.05,d=|0.85|),and a reduced support phase(bilateral:24.31%vs.28.38%,P<0.05,d=|0.88|;non-paretic:66.19%vs.70.13%,P<0.05,d=|0.94|).For early hemiplegia,personal assistant machine-assisted gait training under the suspension protection system helps establish a correct gait pattern,prevents muscle spasms,and improves motor function.展开更多
In this paper a learning mechanism for reactive fuzzy controller design of a mobile robot navigating in unknown environments is proposed. The fuzzy logical controller is constructed based on the kinematics model of a ...In this paper a learning mechanism for reactive fuzzy controller design of a mobile robot navigating in unknown environments is proposed. The fuzzy logical controller is constructed based on the kinematics model of a real robot. The approach to learning the fuzzy rule base by relatively simple and less computational Q-learning is described in detail. After analyzing the credit assignment problem caused by the rules collision, a remedy is presented. Furthermore, time-varying parameters are used to increase the learning speed. Simulation results prove the mechanism can learn fuzzy navigation rules successfully only using scalar reinforcement signal and the rule base learned is proved to be correct and feasible on real robot platforms.展开更多
针对无监督环境下传统网络异常诊断算法存在异常点定位和异常数据分类准确率低等不足,通过设计一种基于改进Q-learning算法的无线网络异常诊断方法:首先基于ADU(Asynchronous Data Unit异步数据单元)单元采集无线网络的数据流,并提取数...针对无监督环境下传统网络异常诊断算法存在异常点定位和异常数据分类准确率低等不足,通过设计一种基于改进Q-learning算法的无线网络异常诊断方法:首先基于ADU(Asynchronous Data Unit异步数据单元)单元采集无线网络的数据流,并提取数据包特征;然后构建Q-learning算法模型探索状态值和奖励值的平衡点,利用SA(Simulated Annealing模拟退火)算法从全局视角对下一时刻状态进行精确识别;最后确定训练样本的联合分布概率,提升输出值的逼近性能以达到平衡探索与代价之间的均衡。测试结果显示:改进Q-learning算法的网络异常定位准确率均值达99.4%,在不同类型网络异常的分类精度和分类效率等方面,也优于三种传统网络异常诊断方法。展开更多
随着智能体在复杂动态环境中的路径规划需求日益增长,传统Q-Learning算法在收敛速度、避障效率及全局优化能力上的局限性逐渐凸显。针对Q-Learning算法在路径规划中的不足,本文提出一种结合动态学习率、自适应探索率与蒙特卡洛树搜索(Mo...随着智能体在复杂动态环境中的路径规划需求日益增长,传统Q-Learning算法在收敛速度、避障效率及全局优化能力上的局限性逐渐凸显。针对Q-Learning算法在路径规划中的不足,本文提出一种结合动态学习率、自适应探索率与蒙特卡洛树搜索(Monte Carlo Tree Search,MCTS)的改进方法。首先,通过引入指数衰减的动态学习率与探索率,以平衡算法在训练初期的探索能力与后期的策略稳定性;其次,将MCTS与Q-Learning结合,利用MCTS的全局搜索特性优化Q值更新过程;此外,融合启发式函数以改进奖励机制,引导智能体更高效地逼近目标。实验结果表明,改进算法的平均步数、收敛速度、稳定性等相较于传统算法提升显著,本研究为复杂环境下的智能体路径规划提供了一种高效、鲁棒的解决方案。展开更多
目的:针对生物安全实验室空间密闭、障碍物形态多(球形、立方体、圆柱体、椭球体等)及精确操作要求极高的复杂环境特性,提出一种融合改进Q-learning和粒子群优化(particle swarm optimization,PSO)算法的机械臂轨迹规划与避障算法QPSO...目的:针对生物安全实验室空间密闭、障碍物形态多(球形、立方体、圆柱体、椭球体等)及精确操作要求极高的复杂环境特性,提出一种融合改进Q-learning和粒子群优化(particle swarm optimization,PSO)算法的机械臂轨迹规划与避障算法QPSO。方法:QPSO算法采用双层优化架构,上层利用改进的Q-learning算法实现路径决策,通过非线性动态温度玻尔兹曼探索策略平衡探索与利用;下层采用含动态权重和学习因子的PSO算法优化轨迹,并结合余弦定理碰撞检测策略保障避障安全性。为验证提出算法的可行性,进行算法性能分析和避障性能测试,并与标准PSO算法、遗传算法、萤火虫算法、改进快速扩展随机树(rapidly-exploring random tree star,RRT*)算法进行对比。结果:相比标准PSO算法、遗传算法、萤火虫算法和RRT*算法,提出的QPSO算法在收敛性能、轨迹长度和避障成功率方面均有显著优势,且在确保最短路径的同时可实现最大安全距离。结论:提出的QPSO算法能有效提升复杂环境下机械臂的轨迹规划和避障效果,可为生物安全实验室等类似环境的自动化实验操作提供可靠的技术支撑。展开更多
文摘BACKGROUND:Hemiplegia,a prevalent stroke-related condition,is often studied for motor dysfunction;however,spasticity remains under-researched.Abnormal muscle tone significantly hinders hemiplegic patients’walking recovery.OBJECTIVE:To determine whether early suspension-protected training with a personal assistant machine for stroke patients enhances walking ability and prevents muscle spasms.METHODS:Thirty-two early-stage stroke patients from Shenzhen University General Hospital and the China Rehabilitation Research Center were randomly assigned to the experimental group(n=16)and the control group(n=16).Both groups underwent 4 weeks of gait training under the suspension protection system for 30 minutes daily,5 days a week.The experimental group used the personal assistant machine during training.Three-dimensional gait analysis(using the Cortex motion capture system),Brunnstrom staging,Fugl-Meyer Assessment for lower limb motor function,Fugl-Meyer balance function,and the modified Ashworth Scale were evaluated within 1 week before the intervention and after 4 weeks of intervention.RESULTS AND CONCLUSION:After the 4-week intervention,all outcome measures showed significant changes in each group.The experimental group had a small but significant increase in the modified Ashworth Scale score(P<0.05,d=|0.15|),while the control group had a large significant increase(P<0.05,d=|1.48|).The experimental group demonstrated greater improvements in walking speed(16.5 to 38.44 cm/s,P<0.05,d=|4.01|),step frequency(46.44 to 64.94 steps/min,P<0.05,d=|2.32|),stride length(15.50 to 29.81 cm,P<0.05,d=|3.44|),and peak hip and knee flexion(d=|1.82|to|2.17|).After treatment,the experimental group showed significantly greater improvements than the control group in walking speed(38.44 vs.26.63 cm/s,P<0.05,d=|2.75|),stride length,peak hip and knee flexion(d=|1.31|to|1.45|),step frequency(64.94 vs.59.38 steps/min,P<0.05,d=|0.85|),and a reduced support phase(bilateral:24.31%vs.28.38%,P<0.05,d=|0.88|;non-paretic:66.19%vs.70.13%,P<0.05,d=|0.94|).For early hemiplegia,personal assistant machine-assisted gait training under the suspension protection system helps establish a correct gait pattern,prevents muscle spasms,and improves motor function.
文摘In this paper a learning mechanism for reactive fuzzy controller design of a mobile robot navigating in unknown environments is proposed. The fuzzy logical controller is constructed based on the kinematics model of a real robot. The approach to learning the fuzzy rule base by relatively simple and less computational Q-learning is described in detail. After analyzing the credit assignment problem caused by the rules collision, a remedy is presented. Furthermore, time-varying parameters are used to increase the learning speed. Simulation results prove the mechanism can learn fuzzy navigation rules successfully only using scalar reinforcement signal and the rule base learned is proved to be correct and feasible on real robot platforms.
文摘针对无监督环境下传统网络异常诊断算法存在异常点定位和异常数据分类准确率低等不足,通过设计一种基于改进Q-learning算法的无线网络异常诊断方法:首先基于ADU(Asynchronous Data Unit异步数据单元)单元采集无线网络的数据流,并提取数据包特征;然后构建Q-learning算法模型探索状态值和奖励值的平衡点,利用SA(Simulated Annealing模拟退火)算法从全局视角对下一时刻状态进行精确识别;最后确定训练样本的联合分布概率,提升输出值的逼近性能以达到平衡探索与代价之间的均衡。测试结果显示:改进Q-learning算法的网络异常定位准确率均值达99.4%,在不同类型网络异常的分类精度和分类效率等方面,也优于三种传统网络异常诊断方法。
文摘随着智能体在复杂动态环境中的路径规划需求日益增长,传统Q-Learning算法在收敛速度、避障效率及全局优化能力上的局限性逐渐凸显。针对Q-Learning算法在路径规划中的不足,本文提出一种结合动态学习率、自适应探索率与蒙特卡洛树搜索(Monte Carlo Tree Search,MCTS)的改进方法。首先,通过引入指数衰减的动态学习率与探索率,以平衡算法在训练初期的探索能力与后期的策略稳定性;其次,将MCTS与Q-Learning结合,利用MCTS的全局搜索特性优化Q值更新过程;此外,融合启发式函数以改进奖励机制,引导智能体更高效地逼近目标。实验结果表明,改进算法的平均步数、收敛速度、稳定性等相较于传统算法提升显著,本研究为复杂环境下的智能体路径规划提供了一种高效、鲁棒的解决方案。
文摘目的:针对生物安全实验室空间密闭、障碍物形态多(球形、立方体、圆柱体、椭球体等)及精确操作要求极高的复杂环境特性,提出一种融合改进Q-learning和粒子群优化(particle swarm optimization,PSO)算法的机械臂轨迹规划与避障算法QPSO。方法:QPSO算法采用双层优化架构,上层利用改进的Q-learning算法实现路径决策,通过非线性动态温度玻尔兹曼探索策略平衡探索与利用;下层采用含动态权重和学习因子的PSO算法优化轨迹,并结合余弦定理碰撞检测策略保障避障安全性。为验证提出算法的可行性,进行算法性能分析和避障性能测试,并与标准PSO算法、遗传算法、萤火虫算法、改进快速扩展随机树(rapidly-exploring random tree star,RRT*)算法进行对比。结果:相比标准PSO算法、遗传算法、萤火虫算法和RRT*算法,提出的QPSO算法在收敛性能、轨迹长度和避障成功率方面均有显著优势,且在确保最短路径的同时可实现最大安全距离。结论:提出的QPSO算法能有效提升复杂环境下机械臂的轨迹规划和避障效果,可为生物安全实验室等类似环境的自动化实验操作提供可靠的技术支撑。
