Iterative learning identification for a class of Wiener nonlinear time-varying systems

Within impedance control framework, since force tracking accuracy is affected by the robotic tracking error, an adaptive force control scheme combined with an iterative learning algorithm is proposed in this paper. First, the unknown environment stiffness and location are observed in real time based on the actual contact force and robot position, and the adaptive reference trajectory is obtained by the estimated environment information. Then, position-based impedance control scheme is adopted to adjust the dynamical relationship between the end-effect or position and the contact force. Considering the effect of the position tracking error on the force tracking accuracy, an iterative learning control method is utilized to compensate the motion error effectively by the force error. The proposed method can effectively reduce the steady state force error by simply combining the indirectly adaptive reference trajectory generation technique with position tracking error regulation method. Stability and convergence conditions are presented with a Lyapunov function and frequency domain method. Simulations and experiments are conducted on a biaxial platform to demonstrate the effectiveness of the proposed method.

Aiming at the problem of a great deal of blindly searching the proportional, differential, integral parameters in the process of designing PID controller, the new algorithm combining genetic algorithm with iterative learning algorithm is named as an iterative learning genetic algorithm (ILGA), which can be used to optimize three controller parameters, thus the optimal parameters can be achieved swiftly by virtue of less iterative learning times, and the design of the PID controller is simplified. As the simulation results shown, the effectiveness of the method is verified in optimizing the PID controller parameters

This paper deals with a new iterative learning dynamic identification method of robot controlled with a Computed Torque Control (CTC) law. The parameters of the Inverse Dynamic Model (IDM) used to compute the CTC, are periodically calculated to minimize the quadratic error between the actual joint force/torque and a joint force/torque calculated with the Inverse Dynamic Identification Model (IDIM), linear in relation to the parameters. Usually the parameters are estimated off-line and the IDIM is calculated with the joint position and its noisy derivatives and cannot take into account on line variations of the parameters (IDIM-LS method). The new method called IDIM-ILIC (IDIM with Iterative Learning Identification and Control) overcomes these 2 drawbacks. The parameters are periodically calculated over a moving time window to update the IDM of the CTC, and the IDIM is calculated with the noise-free data of the trajectory generator, which avoids using the noisy derivatives of the actual joint position. An experimental setup on a prismatic joint validates the procedure with stationary parameters and with a variation of the payload.

This paper discusses a complex-valued Hopfield associative memory with an iterative incremental learning algorithm. The mathematical proofs derive that the weight matrix is approximated as a weight matrix by the complex-valued pseudo inverse algorithm. Furthermore, the minimum number of iterations for the learning sequence is defined with maintaining the network stability. From the result of simulation experiment in terms of memory capacity and noise tolerance, the proposed model has the superior ability than the model with a complex-valued pseudo inverse learning algorithm.

The authors show that a local iterative learning algorithm of Diederich and Opper (1987) converges for any set of patterns, including linearly dependent sets, and that the resulting coupling constants for a given set of patterns are the same as those given by the pseudoinverse rule for any maximal linearly independent subset of the given set. Thus, the matrix of couplings produced by the algorithm is the desired projection matrix onto the subspace spanned by the training set.

This paper presents the development and field validation of an iterative learning-based admittance control algorithm for autonomous excavation in fragmented rock using robotic wheel loaders. An admittance control strategy is augmented with iterative learning, which automatically updates control parameters based on the error between a target bucket fill weight and the measured fill weight at the end of each excavation pass. The algorithm was validated through full-scale autonomous excavation experiments with a 14-tonne capacity load-haul-dump (LHD) machine and two different types of excavation materials: fragmented rock and gravel. In both excavation scenarios, the iterative learning algorithm is able to update the admittance control parameters for a specified target bucket fill weight, eliminating the need to manually re-tune control parameters as material characteristics change. These results have practical significance for increasing the autonomy of robotic wheel loaders used in mining and construction.

In this paper, a new iterative learning algorithm is proposed for repetitive nonlinear systems. The control system employs a combination of state feedback and iterative learning control (ILC) in which the coefficients of states are learned similar to ILC methods. The control system is in a closed loop format both in iteration domain (because of ILC) and in time domain (because of feedback control) which improves the robustness of the conventional ILC. The convergence of the control algorithm is also proved. Finally, simulation results for a 5-DOF manipulator have been presented to illustrate that the proposed algorithm is more robust than a first order P type ILC method at the presence of white Gaussian noise as a nonrepeating disturbance.

Gimbal-less dual axis point-to-point (quasistatic) MEMS mirrors have very wide bandwidths for laser beam steering, however users are often limited to only a third of the bandwidth due to the high Q-factor and use of low-pass filters in open-loop operation to avoid overshoot and oscillation. Closed-loop driving enables the use of the full bandwidth with additional complexity in optics and electronics, which can be undesirable in some low SWaP-C systems. But for many applications which require scanning repetitive patterns, such as LiDAR and biomedical imaging, bandwidth utilization and linearity can be greatly increased without any real-time feedback control by training the device and finding an optimal driving waveform using an iterative learning algorithm. The algorithm drives the device with a trial waveform, measures the scan on a Position Sensing Device (PSD), calculates the error between the desired waveform and the measured position, and adjusts the drive waveform for the next iteration based on an approximate linear device model. This is repeated until the error is reduced to below an acceptable specification. The waveform is then saved in the MEMS Controller and can be reliably used for extended periods of operation. Multiple such drive signals can be trained and stored on the controller to perform different types of scans. Several MEMS mirrors, including single- and dual-axis designs, were studied and three are reported here. Overall, in all cases a high accuracy of optical scans is achieved, typically to within ±0.025° of nominal. Repeatability after training, then running in open loop is better than ±0.01° - however, this measurement was limited by the lower resolution of the position detecting sensor. Scan rates achieved vary based on mirror design, but in each case are greatly improved from those achievable with basic driving approaches. Each mirror demonstrated higher quality vector graphics content at faster refresh rates and stable linear rasters at rates below resonance where lines are scanned with uniform velocity. Additionally, each mirror could achieve stable fast rasters with the line-scanning axis rates just below resonance, giving sinusoidal scans with line rates of ~1.6fres. Finally, each mirror was also demonstrated achieving rasters with rates above resonance, giving line rates of ~2.5fres. In all of those cases the other axis could scan linear and sharp sawtooth or triangle waveforms. Based on the symmetry of the MEMS design, we demonstrated the same performance at ddegerent angles, e.g. rastering at a 45° angle.

Constrained Point-to-Point Iterative Learning Control

Iterative trajectory learning for highly accurate optical satellite tracking systems

In this work, we present a new machine learning-based framework for the accurate estimation of the shot-noise limited photon-counting Poisson channel by developing a state-of-art iterative unsupervised learning algorithm for intelligent optical communications. By accurate estimation, we mean that it achieves very low estimation error even when the effect of the received data is unpredictable. We consider a realistic situation where modulated symbols are assumed to be hidden or unobserved latent variables, thereby making conventional estimation algorithms based on maximum likelihood approach unsuitable or inefficient. In particular, we consider a probabilistic model and assume that the received data is not labeled. With this unpredictable data considered, a novel iterative machine learning framework is developed based on an expectation and maximization algorithm. The proposed algorithm avoids the need to choose an appropriate step size as required in gradient method based algorithms. It is shown that it significantly outperforms the least square and the Viterbi detection technique.

In this paper, we propose and evaluate a simple mechanism to accelerate iterative machine learning algorithms implemented in Hadoop map-reduce (stock), and Apache Spark. In particular, we describe a technique that enables data parallel tasks in map-reduce and Spark to be dynamically and adaptively scheduled on CPU or GPU, based on availability and load. We examine the extent of performance improvements, and correlate them to various parameters of the algorithms studied. We focus on end-to-end performance impact, including overheads associated with transferring data into and out of the GPU, and conversion between data representations in the JVM and on GPU. We also present three optimizations that, in our analysis, can be generalized across many iterative machine learning applications. We present a case study where we accelerate four iterative machine learning applications – multinomial logistic regression, multiple linear regression, K-Means clustering and principal components analysis using singular value decomposition, implemented in three data analytics frameworks – Hadoop Map-Reduce (HMR), IBM Main-Memory Map-Reduce (M3R) and Spark. We observe that the use of GPGPUs decreases the execution time of these applications on HMR by up to 8X, M3R by up to 18X, and Spark by up to 25X. Through our empirical analysis, we offer several insights that can be helpful in designing middleware and cluster managers to accelerate map-reduce and Spark applications using GPUs.

An interpretability improvement for fuzzy rule bases obtained by the iterative rule learning approach

With the increasing deployment of Unmanned Aerial Vehicle (UAV) swarms in uncertain and dynamically changing environments, optimal cooperative control has become essential for ensuring robust and efficient system coordination. To this end, this paper designs a data-driven optimal bipartite containment tracking control scheme for multi-UAV systems under compound uncertainties. A novel Dynamic Iteration Regulation Strategy (DIRS) is proposed, which enables real-time adjustment of the learning iteration step according to the task-specific demands. Compared with conventional fixed-step data-driven algorithms, the DIRS provides greater flexibility and computational efficiency, allowing for better trade-offs between the performance and cost. First, the optimal bipartite containment tracking control problem is formulated, and the associated coupled Hamilton–Jacobi–Bellman (HJB) equations are established. Then, a data-driven iterative policy learning algorithm equipped with the DIRS is developed to solve the optimal control law online. The stability and convergence of the proposed control scheme are rigorously analyzed. Furthermore, the control law is approximated via the neural network framework without requiring full knowledge of the model. Finally, numerical simulations are provided to demonstrate the effectiveness and robustness of the proposed DIRS-based optimal containment tracking scheme for multi-UAV systems, which can reduce the number of iterations by 88.27% compared to that for the conventional algorithm.

This paper aims to investigate the security problem of nonlinear cyber-physical systems (CPSs), which poses a challenge to handle compared with linear CPSs. A series of optimization problems for nonlinear CPSs under injection attack are constructed, which are based on a general model of the nonlinear systems with repetitive operation characteristics and a novel introduction of the key technical lemma. These optimization results are more general than the existing injection attack results and the requirements for attackers to obtain system information are relaxed. Also, the form of switching applied to the attack strategy possesses several advantages, including high stealthiness, lower cost, and more flexibility. Therefore, the new optimal injection attack strategies are expected to be more widespread and provide a basis for the design of defense strategies. The key to acquiring the designed optimal attack strategies is to adopt the linear input/output (I/O) data model for these systems via introducing an estimation term of the improved projection estimation method into the linear model. Finally, a networked GLUON-6L3 manipulator example validates the effectiveness of the proposed methods. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The main purpose of this paper is to study the cyber security of nonlinear cyber-physical systems from the perspective of attackers, which can help defenders fully understand the behavior of attackers. Most of the existing attack strategies aimed at linear systems or have a priori knowledge of the attacked systems. However, there are great limitations and difficulties in practical application. In this paper, the new attack strategies are proposed by combining system identification, iterative learning and control theory, which relaxes the requirement of the attacker’s ability. In detail, the attacker only needs to obtain the control input and output data of the attacked system and design the attack strategy according to it. Among them, the first-in, first-out queue storage method is applied to remove the requirements for the storage capability of the attacker. In practical applications, the attacker does not need to store the I/O data continuously, but only store the initial value and the data of two successive iterations. The mathematical analytic forms of two optimal attack strategies are given, and then their effectiveness are verified on a welding work of a networked six-axis manipulator. In the future, we will investigate the design of attack strategies for nonlinear systems with heterogeneous dynamics and multiple delays.