[目的]为解决非结构化环境下采用深度强化学习进行采摘机械臂路径规划时存在的效率低、采摘路径规划成功率不佳的问题,提出了一种非结构化环境下基于深度强化学习(Deep reinforcement learning, DRL)和人工势场的柑橘采摘机械臂的路径...[目的]为解决非结构化环境下采用深度强化学习进行采摘机械臂路径规划时存在的效率低、采摘路径规划成功率不佳的问题,提出了一种非结构化环境下基于深度强化学习(Deep reinforcement learning, DRL)和人工势场的柑橘采摘机械臂的路径规划方法。[方法]首先,通过强化学习方法进行采摘路径规划问题求解,设计了结合人工势场的强化学习方法;其次,引入长短期记忆(Longshort term memory,LSTM)结构对2种DRL算法的Actor网络和Critic网络进行改进;最后,在3种不同的非结构化柑橘果树环境训练DRL算法对采摘机械臂进行路径规划。[结果]仿真对比试验表明:结合人工势场的强化学习方法有效提高了采摘机械臂路径规划的成功率;引入LSTM结构的方法可使深度确定性策略梯度(Deep deterministic policy gradient,DDPG)算法的收敛速度提升57.25%,路径规划成功率提升23.00%;使软行为评判(Soft actor critic,SAC)算法的收敛速度提升53.73%,路径规划成功率提升9.00%;与传统算法RRT-connect(Rapidly exploring random trees connect)对比,引入LSTM结构的SAC算法使规划路径长度缩短了16.20%,路径规划成功率提升了9.67%。[结论]所提出的路径规划方法在路径规划长度、路径规划成功率方面存在一定优势,可为解决采摘机器人在非结构化环境下的路径规划问题提供参考。展开更多
We propose a new method to generate terahertz perfect vortex beam with integer-order and fractional-order. A new optical diffractive element composed of the phase combination of a spherical harmonic axicon and a spira...We propose a new method to generate terahertz perfect vortex beam with integer-order and fractional-order. A new optical diffractive element composed of the phase combination of a spherical harmonic axicon and a spiral phase plate is designed and called spiral spherical harmonic axicon. A terahertz Gaussian beam passes through the spiral spherical harmonic axicon to generate a terahertz vortex beam. When only the topological charge number carried by spiral spherical harmonic axicon increases, the ring radius of terahertz vortex beam increases slightly, so the beam is shaped into a terahertz quasi-perfect vortex beam. Importantly, the terahertz quasi-perfect vortex beam can carry not only integer-order topological charge number but also fractional-order topological charge number. This is the first time that vortex beam and quasi-perfect vortex beam with fractional-order have been successfully realized in terahertz domain and experiment.展开更多
文摘[目的]为解决非结构化环境下采用深度强化学习进行采摘机械臂路径规划时存在的效率低、采摘路径规划成功率不佳的问题,提出了一种非结构化环境下基于深度强化学习(Deep reinforcement learning, DRL)和人工势场的柑橘采摘机械臂的路径规划方法。[方法]首先,通过强化学习方法进行采摘路径规划问题求解,设计了结合人工势场的强化学习方法;其次,引入长短期记忆(Longshort term memory,LSTM)结构对2种DRL算法的Actor网络和Critic网络进行改进;最后,在3种不同的非结构化柑橘果树环境训练DRL算法对采摘机械臂进行路径规划。[结果]仿真对比试验表明:结合人工势场的强化学习方法有效提高了采摘机械臂路径规划的成功率;引入LSTM结构的方法可使深度确定性策略梯度(Deep deterministic policy gradient,DDPG)算法的收敛速度提升57.25%,路径规划成功率提升23.00%;使软行为评判(Soft actor critic,SAC)算法的收敛速度提升53.73%,路径规划成功率提升9.00%;与传统算法RRT-connect(Rapidly exploring random trees connect)对比,引入LSTM结构的SAC算法使规划路径长度缩短了16.20%,路径规划成功率提升了9.67%。[结论]所提出的路径规划方法在路径规划长度、路径规划成功率方面存在一定优势,可为解决采摘机器人在非结构化环境下的路径规划问题提供参考。
基金Project supported by the Fundamental Research Funds for the Central Universities,China (Grant No.2017KFYXJJ029)。
文摘We propose a new method to generate terahertz perfect vortex beam with integer-order and fractional-order. A new optical diffractive element composed of the phase combination of a spherical harmonic axicon and a spiral phase plate is designed and called spiral spherical harmonic axicon. A terahertz Gaussian beam passes through the spiral spherical harmonic axicon to generate a terahertz vortex beam. When only the topological charge number carried by spiral spherical harmonic axicon increases, the ring radius of terahertz vortex beam increases slightly, so the beam is shaped into a terahertz quasi-perfect vortex beam. Importantly, the terahertz quasi-perfect vortex beam can carry not only integer-order topological charge number but also fractional-order topological charge number. This is the first time that vortex beam and quasi-perfect vortex beam with fractional-order have been successfully realized in terahertz domain and experiment.
