基于自适应学习的序列生成方法

离散序列生成广泛应用于文本生成、序列推荐等领域。目前的研究工作主要集中在提高序列生成的准确性,却忽略了生成的多样性。针对该现象,提出了一种自适应序列生成方法ECoT,设置两层元控制器,在数据层面,使用元控制器实现自适应可学习采样,自动平衡真实数据与生成数据分布得到混合数据分布;在模型层面,添加多样性约束项,并使用元控制器自适应学习最优更新梯度,提升生成模型生成多样性。此外,进一步提出融合协同训练和对抗学习的方法,提升生成模型生成准确性。与目前的主流模型进行对比实验,结果表明,在生成准确性和多样性上,自适应协同训练序列生成方法具有更均衡的准确性和多样性,同时有效缓解生成模型的模式崩溃问题。     

基于多感知的空间机械手爪控制研究

对融合了视觉、滑觉、角位移等多种传感器的欠驱动空间机械手爪,研究其对不同形状、质地的物体实现自适应抓取控制。通过传感器反馈控制机械手运动、抓取力,提高机械手的自主能力。在抓取模式选择中,采用基于专家系统的抓取规划,根据物体不同的形状、尺寸选择不同的抓取模式;在抓取力控制中,通过由PVDF制作的滑觉传感器反馈,采用基于滑觉信号的模糊控制方法,对不同质地的物体选择不同的控制参数。通过实验研究验证基于多感知的控制方法对各种物体可以进行可靠的抓取。     

平夹自适应欠驱动手的参数优化与稳定性分析

针对欠驱动多模式抓取问题,提出了一种具有平夹自适应复合抓取模式的8自由度机械手(PASA手).为了研究PASA手的抓取性能,提出了抓取姿态与稳定性分析模型.根据被抓取物体的位置、尺寸和大小的不同,PASA手能够执行平夹(PA)用于精确抓取,自适应包络用于力量型抓取,或者像人手那样既能平夹又能自适应抓取(PASA).不同抓取模式之间的切换是自动且自适应的.详细讨论了PASA手的运动、抓取力分布和在不同外力下基于势能的抓取稳定性分析.对各种状态下手指接触力进行了测量,并对不同的物体进行了抓取实验.尽管被抓取对象形状、大小各异,PASA手均能牢固抓取.实验结果表明PASA手具有可靠性和广泛适用性.     

杂乱场景下小物体抓取检测研究

目的 杂乱场景下的物体抓取姿态检测是智能机器人的一项基本技能。尽管六自由度抓取学习取得了进展,但先前的方法在采样和学习中忽略了物体尺寸差异,导致在小物体上抓取表现较差。方法 提出了一种物体掩码辅助采样方法,在所有物体上采样相同的点以平衡抓取分布,解决了采样点分布不均匀问题。此外,学习时采用多尺度学习策略,在物体部分点云上使用多尺度圆柱分组以提升局部几何表示能力,解决了由物体尺度差异导致的学习抓取操作参数困难问题。通过设计一个端到端的抓取网络,嵌入了提出的采样和学习方法,能够有效提升物体抓取检测性能。结果 在大型基准数据集GraspNet-1Billion上进行评估,本文方法取得对比方法中的最优性能,其中在小物体上的抓取指标平均提升了7%,大量的真实机器人实验也表明该方法具有抓取未知物体的良好泛化性能。结论 本文聚焦于小物体上的抓取,提出了一种掩码辅助采样方法嵌入到提出的端到端学习网络中,并引入了多尺度分组学习策略提高物体的局部几何表示,能够有效提升在小尺寸物体上的抓取质量,并在所有物体上的抓取评估结果都超过了对比方法。     

基于点云采样权重估计的未知物体抓取位姿生成方法

针对机械臂在非结构环境中对未知物体抓取位姿生成困难及抓取稳定性差的问题,提出一种基于点云采样权重估计的抓取位姿生成方法.首先通过移动深度相机的方式拼接得到较完整的物体点云信息,并对物体的几何特性进行分析,有效避开物体不宜抓取的位置进行抓取位姿样本生成;然后结合几何约束条件实现抓取位姿搜索,并利用力封闭条件对样本稳定性进行评估;最后为了对实际的抓取位姿进行评价,根据其稳定性、夹取深度、夹取角度等设定抓取可行性指标,据此在工作空间输出最佳抓取位姿并完成指定的抓取任务.实验结果表明,采用所提方法能够高效生成大量且稳定的抓取位姿,并在仿真环境中有效实现机械臂对单个或多个随机摆放的未知物体的抓取任务.     

形状自适应欠驱动三关节机器人手指设计

根据欠驱动原理研制的三指10个自由度的机器人手爪具有驱动元件数量少、抓取物体范围广泛等优点．在欠驱动手爪的4个主要机构中,欠驱动手指对抓取物体具有被动柔顺和形状自适应的特性．首先对三关节欠驱动手指机构进行静力学分析,提出合理的设计目标和约束条件;然后根据设计目标,采用遗传算法得到手指机构的各个关节连杆尺寸和抓取物体时的特殊构形,使得在抓取给定物体时各关节指面的接触力达到均匀分布,得到高效的力传递和更加紧凑的机构尺寸．     

差动瓦特连杆直线平夹自适应抓持器

现有的平夹自适应抓持器不能实现末端直线轨迹运动,导致其存在抓取盲区。针对此问题,本文提出一种新型平夹自适应手指——Watt手指。该手指采用瓦特连杆机构实现末端指段的直线轨迹运动,结合平行四连杆机构保持末端指段在初始阶段的固定姿态,并利用三角形差动连杆机构实现自适应抓取。Watt手指在初始阶段具有直线平行夹持功能,适合捏持桌面上不同尺寸的物体;当第1或第2指段接触物体被阻挡时,该装置自动进入自适应抓取模式,可以适应不同形状、尺寸的物体。对该Watt手指进行了运动学分析,获得了该手指的稳定抓取空间和相对优化的参数集合。研制了二指Watt手,理论分析与实验结果表明,Watt手可实现良好的直线平夹与自适应抓取效果,直线轨迹使得该手具有更大的抓取空间,适应范围更广。     

机器人多指操作的递阶控制

为机器人多指协调操作建立一递阶控制系统.给定一操作任务,任务规划器首先生 成一系列物体的运动速度;然后,协调运动规划器根据期望的物体运动速度生成期望的手指 运动速度和期望的抓取姿态变化;同时,抓取力规划器为平衡作用在物体上的外力,根据当前 的抓取姿态,生成各手指所需的抓取力;最后,系统将手指的期望运动速度与为实现期望抓取 力而生成的顺应速度合并,并通过手指的逆雅可比转化为手指关节运动速度后,由手指的关 节级运动控制器实现手指的运动和抓取力的控制.该控制方法已成功应用于香港科技大学 (HKUST)灵巧手控制系统的开发.实验证明该方法不仅能完成物体轨迹的跟踪控制任务, 而且能完成物体对环境的力控制和力与速度的混合控制.     

基于视觉感知和触觉先验知识学习的机器人稳定抓取

针对机器人抓取过程中需要实时评估抓取质量以动态调整抓取构型的问题,提出了基于触觉先验知识的机器人稳定抓取方法.首先,根据抓取过程中物体抵抗外界扰动的能力,提出了一种基于触觉信息的抓取质量评估方法.在此基础上,建立了视触觉联合数据集并学习触觉先验知识.其次,提出了融合视觉图像和触觉先验知识的稳定抓取构型生成方法.最后,在搭建的机器人抓取系统中对10种目标物体进行了实验验证.结果表明,相比传统的视觉方法,该方法的抓取稳定性提高了55％;针对已知物体和未知物体,稳定抓取成功率分别为86％和79％,体现了较好的泛化效果.     

虚拟手自然抓取动作生成研究

在虚拟抓取规划中,生成稳定的抓取姿态非常重要,但是只考虑稳定性往往会导致抓取姿态不符合自然的人手抓取习惯.为了生成既满足物理约束又符合人手抓取习惯的自然抓取姿态,提出一种基于抓取分类理论的虚拟手自然抓取规划算法.首先根据人类抓取分类理论,基于人类行为构建物体形状与人手抓取姿态之间的映射关系,然后在动作规划中加入人手抓取姿态的引导,最终在虚拟手的位姿空间中找到一组自然的抓取姿态.对于复杂的多部件物体,先将其分割成可抓的部件,再对可抓部件进行自然抓取动作的生成.实验中展示了以人类抓取分类理论中的5种人手抓取姿态为指导,对常见的三维物体生成自然抓取姿态的结果,经用户评估,所提算法生成的抓取动作相比其他算法更符合人手抓取习惯.     

A new linear adaptive controller: design, analysis and performance

The certainty equivalence and polynomial approach, widely used for designing adaptive controllers, leads to “simple” adaptive control designs that guarantee stability, asymptotic error convergence, and robustness, but not necessarily good transient performance. Backstepping and tuning functions techniques, on the other hand, are used to design adaptive controllers that guarantee stability and good transient performance at the expense of a highly nonlinear controller. In this paper, we use elements from both design approaches to develop a new certainty equivalence based adaptive controller by combining backstepping based control law with a normalized adaptive law. The new adaptive controller guarantees stability and performance, as well as parametric robustness for the nonadaptive controller, that are comparable with the tuning functions scheme, without the use of higher order nonlinearities     

仿生假手抓握力控制策略

为了使仿生假手完成各种精细作业,提出一种抓握力控制策略.在自由空间和约束空间中分别使用基于位置的阻抗控制和力跟踪阻抗控制.在过渡过程中使用模糊观测器切换控制模式.两种控制模式采用同一个基于位置的阻抗控制器,在约束空间向阻抗控制器中引入参考力,以满足约束空间的抓握力控制要求.这种方法可以使关节在自由空间和约束空间中分别实现良好的轨迹跟踪和力矩跟踪,在过渡过程中实现控制模式的可靠切换和系统的稳定过渡.提出一种自适应滑模摩擦力补偿方法,利用终端滑模思想设计了滑模函数,使得系统跟踪误差在有限时间内收敛,避免了传统线性滑模面状态跟踪误差无法在有限时间内收敛至0的问题.根据指数形式摩擦力的特点,利用终端滑模控制思想获得包含摩擦力参数估计的滑模控制律,并基于李亚普诺夫稳定性定理推导了估计参数的在线自适应律.对该抓握力控制策略在HIT假手上进行了抓取实验,实验结果证明了控制策略的有效性.     

非线性系统的直接自适应输出反馈监督模糊控制

针对一类单输入单输出非线性不确定系统，提出一种稳定的直接自适应模糊输出反馈监督控制算法，该算法不需要系统的状态完全可测的假设条件，监督控制不仅迫使系统的状态在指定的集合内，而且当模糊自适应控制处于良好的工作状态时，监督控制可以关闭，证明了整个模糊自适应输出反馈控制算法可以保证闭环系统稳定。     

Direct adaptive fuzzy-neural control with state observer andsupervisory controller for unknown nonlinear dynamical systems

In this paper, an observer-based direct adaptive fuzzy-neural network (FNN) controller with supervisory mode for a certain class of high order unknown nonlinear dynamical system is presented. The direct adaptive control (DAC) has the advantage of less design effort by not using FNN to model the plant. By using an observer-based output feedback control law and adaptive law, the free parameters of the adaptive FNN controller can be tuned on-line based on the Lyapunov synthesis approach. A supervisory controller is appended into the FNN controller to force the state to be within the constraint set. Therefore, if the FNN controller cannot maintain the stability, the supervisory controller starts working to guarantee stability. On the other hand, if the FNN controller works well, the supervisory controller will be de-activated. The overall adaptive scheme guarantees the global stability of the resulting closed-loop system in the sense that all signals involved are uniformly bounded. Simulation results also show that our initial control effort is much less than those in previous works, while preserving the tracking performance     

Adaptive hybrid intelligent control for uncertain nonlinear dynamical systems

A new hybrid direct/indirect adaptive fuzzy neural network (FNN) controller with a state observer and supervisory controller for a class of uncertain nonlinear dynamic systems is developed in this paper. The hybrid adaptive FNN controller, the free parameters of which can be tuned on-line by an observer-based output feedback control law and adaptive law, is a combination of direct and indirect adaptive FNN controllers. A weighting factor, which can be adjusted by the tradeoff between plant knowledge and control knowledge, is adopted to sum together the control efforts from indirect adaptive FNN controller and direct adaptive FNN controller. Furthermore, a supervisory controller is appended into the FNN controller to force the state to be within the constraint set. Therefore, if the FNN controller cannot maintain the stability, the supervisory controller starts working to guarantee stability. On the other hand, if the FNN controller works well, the supervisory controller will be deactivated. The overall adaptive scheme guarantees the global stability of the resulting closed-loop system in the sense that all signals involved are uniformly bounded. Two nonlinear systems, namely, inverted pendulum system and Chua's (1989) chaotic circuit, are fully illustrated to track sinusoidal signals. The resulting hybrid direct/indirect FNN control systems show better performances, i.e., tracking error and control effort can be made smaller and it is more flexible during the design process.     

Robust object manipulation for tactile-based blind grasping

Tactile-based blind grasping addresses realistic robotic grasping in which the hand only has access to proprioceptive and tactile sensors. The robotic hand has no prior knowledge of the object/grasp properties, such as object weight, inertia, and shape. There exists no manipulation controller that rigorously guarantees object manipulation in such a setting. Here, a robust control law is proposed for object manipulation in tactile-based blind grasping. The analysis ensures semi-global asymptotic and exponential stability in the presence of model uncertainties and external disturbances that are neglected in related work. Simulation and hardware results validate the effectiveness of the proposed approach.     

Observer-based adaptive control of uncertain time-delay switched systems with stuck actuator faults

This paper concerns the observer-based adaptive control problem of uncertain time-delay switched systems with stuck actuator faults. Under the case where the original controller cannot stabilize the faulty system, multiple adaptive controllers are designed and a suitable switching logic is incorporated to ensure the closed-loop system stability and state tracking. New delay-independent sufficient conditions for asymptotic stability are obtained in terms of linear matrix inequalities based on piecewise Lyapunov stability theory. On the other hand, adaptive laws for on-line updating of some of the controller parameters are also designed to compensate the effect of stuck failures. Finally, simulation results for reference [1] model show that the design is feasible and efficient.     

Human Model Reference Adaptive Control of a Prosthetic Hand

A human model reference adaptive controller (HMRAC) is developed for a prosthetic hand. The model reference for the adaptive controller is formed from grasp experiments with human test subjects. This HMRAC is incorporated within a hybrid force-position control law; electromyogram (EMG) signals from amputee and nonamputee test subjects are used to control the force or position of the prosthetic hand. The HMRAC is compared to a sliding mode controller (SMC) with high and low control gains during bench top experiments with step and EMG inputs while grasping high and low stiffness objects. Results from the bench top experiments show that the SMC with a high control gain produced the least amount of tracking error with the EMG input at the expense of a highly oscillatory system response while grasping the high stiffness object. The HMRAC produced less tracking error with the step inputs in all cases and less tracking error with the EMG input in comparison to the SMC with a low control gain. The HMRAC also produced less percent overshoot (OS) with the step inputs on average in comparison to the SMC in all cases. Experiments were also performed by a transradial amputee with the SMC and the HMRAC. Both controllers were compared to the amputee’s current prosthesis for daily use. Results from the experiments performed by the amputee with the HMRAC and the SMC were similar to the bench top experiments: the high gain SMC had the least tracking error on average at the expense of a highly oscillatory system response with high object stiffness. The HMRAC was not oscillatory and had the next lowest amount of tracking error than all other prosthesis control options. The HMRAC had slightly more error than the amputee had while using his natural left hand. Similar results were obtained from seven nonamputees who also participated in this study.     

汽车转向/防抱死制动协同控制

为了解决汽车转向过程中防抱死制动稳定性问题,提出一种新的协同控制系统.该协同控制结构由转向控制器和制动控制器组成.在转向控制中设计滑模鲁棒自适应控制器和横摆力矩控制器力求改善汽车动态响应,鲁棒自适应性和稳定性.此外定义协同误差,建立汽车协同误差模型并设计汽车防抱死制动鲁棒自适应控制系统.为了减少转向系统和制动系统之间的补偿控制律难以确定的困难,提出耦合误差补偿原理与同一给定控制相结合的新的耦合控制策略.最后用仿真结果验证所设计控制算法的有效性.