Feedforward Approach Research Articles

An Improved Collaborative Control Scheme to Resist Grid Voltage Unbalance for BDFG-Based Wind Turbine

This article presents an improved collaborative control to resist grid voltage unbalance for brushless doubly fed generator (BDFG)-based wind turbine (BDFGWT). The mathematical model of grid-connected BDFG including machine side converter (MSC) and grid side converter (GSC) in the αβ reference frame during unbalanced grid voltage condition is established. On this base, the improved collaborative control between MSC and GSC is presented. Under the control, the control objective of the whole BDFGWT system, including canceling the pulsations of electromagnetic torque and the unbalance of BDFGWT’s total currents, pulsations of BDFGWT’s total powers are capable of being realized; therefore, the control capability of BDFGWT to resist unbalanced grid voltage is greatly improved. Moreover, improved single-loop current controllers adopting PR regulators are proposed for both MSC and GSC where the sequence extractions for both MSC and GSC currents are not needed any more, and hence the proposed control is much simpler. In addition, the transient characteristics are also improved. Moreover, in order to achieve the decoupling control of current and average power, current controller also adopts the feedforward control approach. Case studies for a two MW BDFGWT system are implemented, and the results verify that the presented control is capable of effectively improving the control capability for BDFGWT to resist grid voltage unbalance and exhibit good stable and dynamic control performances.

Linearizing AFM raster scans acquired using dual-stage lateral scanners

Dual-stage scanners can successfully increase the range of atomic force microscope (AFM) scanners in space and/or frequency by designing and constructing two nanopositioners with significant differences in static and dynamic characteristics. In such cases, the positioning signal has to be split to meet the displacement and frequency limitations of each stage. Without closed-loop displacement measurement sensors, linearizing images acquired using signals split in time and frequency can be challenging. A common method, often cumbersome and time consuming, uses inverse model-based feedforward compensation. In this work, we show that image-based linearization can be used to achieve acceptable results for raster scans acquired using dual-stage scanners. In particular, we apply the feedforward and image-based methods to our homemade dual-stage lateral scanner and high-speed AFM (HS-AFM) system. The acquired scans compare well for the two methods, indicating that image-based raster scan linearization can be used in place of inverse model-based feedforward approaches.

A MATLAB-based simulator for the study of process control of fed-batch yeast fermentations

This paper presents a MATLAB-based simulator for the study of process control of fed-batch yeast fermentations that meets the educational needs of undergraduate and graduate students. Against the background of challenges in interdisciplinary education, the paper examines the evolving environment of simulation tools in the field of bioprocesses. It emphasizes the need for interdisciplinarity to prepare students for the complexities of the modern biotechnology industry. Built with accessibility and flexibility in mind, the simulator offers a modular structure with a graphical user interface (GUI) for novice users and direct access to MATLAB functions for advanced users. This design choice ensures ease of use for students with different programming backgrounds and allows for adaptability to alternative software environments. The introductory sections provide a detailed overview of the simulator development, including the mathematical models that govern biomass growth kinetics, mass transfer, and process quality indicators. Selected process control strategies such as rule-based and feedforward approaches are incorporated in the simulator to enhance the learning experience by allowing students to experiment with different scenarios. The paper concludes by emphasizing the simulator's adaptability, modularity, and user-friendly interface as a valuable asset in educating students about the complexities of bioprocess control.

Analysis and evaluation of two reference LiDAR-assisted control designs for wind turbines

LiDAR-assisted wind turbine control holds promise in reducing structural loads and enhancing rotor speed regulation. However, a research gap exists in the practicality and limitations of commercially available fixed-beam LiDARs for large turbines and evaluating commonly employed LiDAR-assisted feedforward approaches. This study addresses these gaps by examining the implications of utilizing fixed-beam LiDARs in two wind turbine sizes and two reference LiDAR-assisted control strategies. A comprehensive evaluation considers coherence variations, uncertainties related to inaccurate pitch angle mapping with the upcoming wind speed, and their combined impact on load reduction. Numerical simulations reveal that an excessively low cut-off frequency in the low-pass filter can compromise preview time compensation. This is problematic in larger turbines, where coherence with limited LiDAR beams is inferior compared to smaller wind turbines, which deteriorates the effectiveness of the LiDAR-assisted control. Among the reference LiDAR-assisted control methods, the evaluation indicates the Schlipf approach has greater load reduction independence, while Bossanyi’s approach, which uses measurement of current blade pitch, yields positive results with fine-tuned baseline controllers. However, allowing baseline controller-induced frequencies to propagate into the controller may increase system excitation at certain frequencies due to the use of the actual pitch angle for feedforward pitch rate calculation.

Artificial Neural Networks and Experimental Analysis of the Resistance Spot Welding Parameters Effect on the Welded Joint Quality of AISI 304.

The automobile industry relies primarily on spot welding operations, particularly resistance spot welding (RSW). The performance and durability of the resistance spot-welded joints are significantly impacted by the welding quality outputs, such as the shear force, nugget diameter, failure mode, and the hardness of the welded joints. In light of this, the present study sought to determine how the aforementioned welding quality outputs of 0.5 and 1 mm thick austenitic stainless steel AISI 304 were affected by RSW parameters, such as welding current, welding time, pressure, holding time, squeezing time, and pulse welding. In order to guarantee precise evaluation and experimental analysis, it is essential that they are supported by a numerical model using an intelligent model. The primary objective of this research is to develop and enhance an intelligent model employing artificial neural network (ANN) models. This model aims to provide deeper knowledge of how the RSW parameters affect the quality of optimum joint behavior. The proposed neural network (NN) models were executed using different ANN structures with various training and transfer functions based on the feedforward backpropagation approach to find the optimal model. The performance of the ANN models was evaluated in accordance with validation metrics, like the mean squared error (MSE) and correlation coefficient (R2). Assessing the experimental findings revealed the maximum shear force and nugget diameter emerged to be 8.6 kN and 5.4 mm for the case of 1-1 mm, 3.298 kN and 4.1 mm for the case of 0.5-0.5 mm, and 4.031 kN and 4.9 mm for the case of 0.5-1 mm. Based on the results of the Pareto charts generated by the Minitab program, the most important parameter for the 1-1 mm case was the welding current; for the 0.5-0.5 mm case, it was pulse welding; and for the 0.5-1 mm case, it was holding time. When looking at the hardness results, it is clear that the nugget zone is much higher than the heat-affected zone (HZ) and base metal (BM) in all three cases. The ANN models showed that the one-output shear force model gave the best prediction, relating to the highest R and the lowest MSE compared to the one-output nugget diameter model and two-output structure. However, the Levenberg-Marquardt backpropagation (Trainlm) training function with the log sigmoid transfer function recorded the best prediction results of both ANN structures.

A New Feed-Forward Control for Dynamics Improvement in a Dual-Input DC–DC Converter for Hybrid Vehicle Applications

In this study, a double-input single-output bidirectional DC–DC converter is considered. This particular architecture allows less switches to be used than a conventional solution. A new feed-forward current control for this DC–DC converter with three switches is presented in this paper. The modulation technique proposed in the literature for the aforementioned converter leads to a consistent loss reduction at low load, exploiting the DCM. As a drawback, when using this control strategy, the dynamic response worsens significantly. To speed up the control, a feed-forward approach is designed and implemented using a simplified converter electrical model. The proposed strategy is compared with the conventional PI controller, and it is validated and verified through simulation results in the MATLAB/Simulink/PLECS environment and through experimental tests using a converter prototype.

Research on adaptive feedforward control method for Tiltrotor Aircraft/Turboshaft engine system based on radial basis function neural network

The paper proposes an adaptive feedforward control approach based on a radial basis function neural network to address the issue of poor disturbance rejection capability in conventional integrated control methods for tiltrotor aircraft propulsion systems. Firstly, a comprehensive simulation model of the tiltrotor aircraft and turboshaft engine is established, and its accuracy is validated against the results of the General Tiltrotor Integrated Simulation Code. Building upon this, a cascade Proportional-integral control parameter design method based on separation principles is proposed to tackle the problem of designing control parameters for the Proportional Integral main loop structure. Subsequently, an adaptive feedforward control method that takes into account both aircraft-induced disturbances and model uncertainties for turboshaft engines is presented using a Radial Basis Function neural network. This approach includes a nonlinear estimator capable of estimating load disturbances and a Radial Basis Function adaptive feedforward controller. Furthermore, stability proof of the closed-loop system is provided based on Lyapunov function analysis. Finally, digital simulations and hardware-in-the-loop experiments are conducted based on a typical tiltrotor flight mission. The results demonstrate that the proposed method effectively mitigates the power turbine drop compared to the conventional Proportional Integral controller with collective pitch feedforward.

Artificial Neural Network Prediction Model for Maternal Health Services Quality in Nigeria

This study presents an innovative approach in predicting the quality of maternal health services in Nigeria using artificial neural networks (ANNs). It focuses on a binary classification task in categorizing maternal health services facilities into "High_Quality" and "Low_Quality" classes. Dataset was collected from diverse healthcare facilities in Nigeria through a WHO Health Resources and Services Availability Monitoring System (HeRAMS) cross sectional-survey report. The collected data was preprocessed; the ANN architecture was developed and trained using back-propagation method. The network was modeled using feed-forward modeling approach with 16 variables at the input layer, 2 hidden layers of 65 and 55 neurons each, and an output of 2 neurons. Hence, 87.17% accuracy with only 2.48% deviation at the 50 epoch was achieved. The model's performance was evaluated using standard metrics of accuracy (87%), precision (88%), recall (89%), and F1-score (88%) to assess its ability to accurately classify maternal health services facilities. The findings of the study can have significant implications to policymakers, healthcare providers, and maternal health advocates to optimize healthcare resources and achieve better maternal health outcomes in Nigeria. Data availability, potential biases in the data, and the subjective nature of some quality-related indicators were acknowledged as limitations. However, despite these challenges, the ANN classification model remains a valuable tool for supporting decision-making in maternal healthcare facility delivery service in Nigeria. Artificial neural network, Binary classification, Health services, Maternal health, Predictive modeling

Inter-discharge optimization for fast, reliable access to ASDEX Upgrade advanced tokamak scenario

Rapid inter-discharge simulation and optimization using the RAPTOR code have allowed the development of a reliable and reproducible early heating strategy for an advanced tokamak (AT) scenario on ASDEX Upgrade. Solving for electron heat and current diffusion in RAPTOR with ad-hoc formulas for heat transport and electron cyclotron current drive (ECCD) efficiency is found to robustly recover the coupled dynamics of Te and q, while maintaining model parameters fixed for all discharges. The pedestal top boundary condition in pre-shot simulations is set by a newly derived scaling law for the electron pressure at ρ = 0.8, using a data set of previous AT discharges. RAPTOR simulations have allowed to develop an understanding of the onset of 3/2 tearing modes, which were observed to have a detrimental impact on confinement when low magnetic shear conditions are present at the rational surface during the high-β phase. Delaying the NBI heating, by a specific time interval found via simulations, has led to avoiding these modes. A non-linear optimization scheme has been applied to optimize the ECCD deposition radii to reach a stationary state with qmin>1 at the beginning of the flat-top phase, while ensuring a non-zero magnetic shear at q = 1.5 throughout the high-β phase, and has been successfully tested in experiment. However, further experiments, aiming for qmin>1.5 , have highlighted limitations of the present feedforward control approach in the presence of shot-to-shot variations that are not included in the applied model. Application of real-time model-based control is proposed to overcome model-reality mismatches in future work.

LIDAR‐assisted feedforward individual pitch control of a 15 MW floating offshore wind turbine

AbstractNacelle‐mounted, forward‐facing light detection and ranging (LIDAR) technology can deliver benefits to rotor speed regulation and loading reductions of floating offshore wind turbines (FOWTs) when assisting with blade pitch control in above‐rated wind speed conditions. Large‐scale wind turbines may be subject to significant variations in structural loads due to differences in the wind profile across the rotor‐swept area. These loading fluctuations can be mitigated by individual pitch control (IPC). This paper presents a novel LIDAR‐assisted feedforward IPC approach that uses each blade's rotor azimuth position to allocate an individual pitch command from a multi‐beam LIDAR. In this study, the source code of OpenFAST wind turbine modelling software was modified to enable LIDAR simulation and LIDAR‐assisted control. The LIDAR simulation modifications were accepted by the National Renewable Energy Laboratory (NREL) and are now present within OpenFAST releases from v3.5 onwards. Simulations of a 15 MW FOWT were performed across the above‐rated wind spectrum. Under a turbulent wind field with an average wind speed of 17 ms−1, the LIDAR‐assisted feedforward IPC delivered up to 54% reductions in the root mean squared errors and standard deviations of key FOWT parameters. Feedforward IPC delivered enhancements of up to 12% over feedforward collective pitch control, relative to the baseline feedback controller. The reductions to the standard deviation and range of the rotor speed may enable structural optimization of the tower, while the reductions in the variations of the loadings present an opportunity for reduced fatigue damage on turbine components and, consequently, a reduction in maintenance expenditure.

On analysis of complex administrative data: neural networks, modelling and prediction

Eco-geographical heterogenicity of countries such as Chile and in cities exhibiting a territorial and demographic important diversity is relevant for the epidemiological studies of the apparition and spread of SARS-CoV-2. That situation is the opposite to countries such as the Czech Republic, with small or less diverse territories where the apparition and spread of SARS-CoV-2 can be correlated mainly to demographic and seasonal variables more than climate, pollution, or other physical and biological variables. It is well visible that there is no simple model for measured active cases and given parameters. This motivates a more general question to develop models that predict the future number of reported COVID-19 cases (by laboratory testing). We tune several neural networks to show the complexity issues of such a problem. These predictions are made for the countries, Austria, Czech Republic, and Slovakia, with the model class used for making these predictions being the artificial neural network, for the data from February 2020 until February 2021. Two different architectures of the neural network are compared the feed-forward network and the recurrent neural network. Ultimately, it is found that there are notable differences between the three countries studied, with the data for the Czech Republic being easier to predict with good accuracy than the data from the other two countries. Likewise, it turns out that the feed-forward approach delivers better results for Austria and the Czech Republic, whereas, for Slovakia, the recurrent approach performs better. Likewise, it is found that combining the data from all three countries does not lead to improved accuracy compared to models using data from only one single country. Both of the findings mentioned above might be related to the relatively small amount of data available.

Combining a forward supervised filter learning with a sparse NMF for breast cancer histopathological image classification

Histopathological images play a important role in clinical diagnosis, particularly in identifying and assessing the severity of abnormal conditions like benign lesions and malignant tumors. Traditional machine learning techniques for processing histopathology images involve the extraction of manual features from these images, which is typically done with the assistance of industry experts. Recent advancements in Deep Learning (DL), especially with Convolutional Neural Networks (CNN), have enabled the automatic extraction of multi-level abstract features directly from raw data. This capability significantly enhances the performance of complex computer vision tasks. Classic CNN models like AlexNet and VggNet employ back-propagation algorithms to learn filters in the training phase. However, these algorithms demand large labeled datasets, resulting in extensive computational processing. Additionally, they often face the vanishing gradient problem, which can negatively impact the quality of the learning process. Besides, in many domains, acquiring enough labeled images for conducting properly the training phase is a real challenge. To address these challenges, a feed-forward propagation approach was proposed using Non-Negative Matrix Factorization(NMF). The NMF technique factorizes the input data into two latent factors (non-negative matrices). It has been shown that by enforcing constraints such as sparsity on the latent factors, dominant features that are mostly correlated with tumors types can be extracted. In this work, a novel model combining sparse NMF and Support Vector Machine (SVM) was developed for classifying histopathological images. We have derived a mathematical model of a novel feed-forward filter learning approach that combines sparse NMF (SNMF) and Support Vector Machine technique (SVM). The model was used to design and implement a feed-forward CNN classifier to classify histopathology images. This model has been evaluated on the histopathology images from Sultan Qaboos University Hospital (SQUH dataset) and the public BreaKHis dataset. The experiments we have conducted demonstrate the efficiency of the proposed model, especially on small-sized SQUH datasets achieving an AUC of 0.90, 0.89, 0.85, and 0.86 on 4x,10x, 20x, and 40x magnifications, respectively, and achieving an AUC of 0.95 BreaKHis dataset.

A data-driven approach to guide supersonic impinging jet control

A data-driven framework using snapshots of an uncontrolled flow is proposed to identify, and subsequently demonstrate, effective control strategies for different objectives in supersonic impinging jets. The open-loop, feed-forward control approach, based on a dynamic mode decomposition reduced-order model (DMD-ROM), computes forcing receptivity in an economical manner by projecting flow and actuator-specific forcing snapshots onto a reduced subspace and then evolving the dynamics forwards in time. Since it effectively determines a linear response around the unsteady flow in the time domain, the method differs materially from typical techniques that use steady basic states, such as stability or input–output approaches that employ linearized Navier–Stokes operators in the frequency domain. The method presented naturally accounts for factors inherent to the snapshot basis, including configuration complexity and flow parameters such as Reynolds number. Furthermore, gain metrics calculated in the reduced subspace facilitate rapid assessments of flow sensitivities to a wide range of forcing parameters, from which optimal actuator inputs may be selected and results confirmed in scale-resolved simulations or experiments. The DMD-ROM approach is demonstrated from two different perspectives. The first concerns asymptotic feedback resonance, where the effects of harmonic pressure forcing are estimated and verified with nonlinear simulations using a blowing–suction actuator. The second examines time-local behaviour within critical feedback events, where the phase of actuation becomes important. For this, a conditional space–time mode is used to identify the optimal forcing phase that minimizes convective instability growth within the resonance cycle.

Feed-forward networks using logistic regression and support vector machine for whole-slide breast cancer histopathology image classification

The performance of an image classification depends on the efficiency of the feature learning process. This process is a challenging task that traditionally requires prior knowledge from domain experts. Recently, representation learning was introduced to extract features directly from the raw images without any prior knowledge. Deep learning using a Convolutional Neural Network (CNN) has gained massive attention for performing image classification, as it achieves remarkable accuracy that sometimes exceeds human performance. But this type of network learns features by using a back-propagation approach. This approach requires a huge amount of training data and suffers from the vanishing gradient problem that deteriorates the feature learning. The forward-propagation approach uses predefined filters or filters learned outside the model and applied in a feed-forward manner. This approach is proven to achieve good results with small size labeled datasets. In this work, we investigate the suitability of using two feed-forward methods such as Convolutional Logistic Regression Network (CLR), and Convolutional Support Vector Machine Network for Histopathology Images (CSVM-H). The experiments we have conducted on two small breast cancer datasets (Sultan Qaboos University Hospital (SQUH) and BreaKHis dataset) demonstrate the advantage of using feed-forward approaches over the traditional back-propagation ones. On those datasets, the proposed models CLR and CSVM-H were faster to train and achieved better classification performance than the traditional back-propagation methods (VggNet-16 and ResNet-50) on the SQUH dataset. Importantly, our proposed approach CLR and CSVM-H efficiently learn representations from small amounts of breast cancer whole-slide images and achieve an AUC of 0.83 and 0.84, respectively, on the SQUH dataset. Moreover, the proposed models reduce memory footprint in the classification of Whole-Slide histopathology images since their training time is significantly reduced compared to the traditional CNN on the SQUH and BreaKHis datasets.

High-Precision Control of Industrial Robot Manipulator Based on Extended Flexible Joint Model

High-precision industrial manipulators are essential components in advanced manufacturing. Model-based feedforward is the key to realizing the high-precision control of industrial robot manipulators. However, traditional feedforward control approaches are based on rigid models or flexible joint models which neglect the elasticities out of the rotational directions and degrade the setpoint precision significantly. To eliminate the effects of elasticities in all directions, a high-precision setpoint feedforward control method is proposed based on the output redefinition of the extended flexible joint model (EFJM). First, the flexible industrial robots are modeled by the EFJM to describe the elasticities in joint rotational directions and out of the rotational directions. Second, the nonminimum-phase EFJM is transformed into a minimum-phase system by using output redefinition. Third, the setpoint control task is transformed from Cartesian space into joint space by trajectory planning based on the EFJM. Third, a universal recursive algorithm is designed to compute the feedforward torque based on the EFJM. Moreover, the computational performance is improved. By compensating the pose errors caused by elasticities in all directions, the proposed method can effectively improve the setpoint control precision. The effectiveness of the proposed method is illustrated by simulation and experimental studies. The experimental results show that the proposed method reduces position errors by more than 65% and the orientation errors by more than 62%.

Comparative analysis of machine learning algorithms on prediction of the sound absorption coefficient for reconfigurable acoustic meta-absorbers

Exploring the potential of sound absorption is foremost for improving room acoustics and unleashing the power of noise remediation. Acoustic metamaterials are considered promising candidates for tunable broadband sound absorption in the low-to-high frequency range. In this work, an acoustic meta-absorber with multi-order Helmholtz resonator cavities is proposed. The meta-absorber unit cell comprises multiple disks, spacers, and a bottom cavity arranged in a series configuration. Therefore, the assembled acoustic meta-absorber is reconfigurable to have different absorption characteristics. Experimental results demonstrate that the proposed meta-absorber unit cell exhibits high sound absorption, with multiple absorption peaks observed within the 80-1600 Hz frequency range. Additionally, tunable broadband and quasi-perfect absorption (≥0.9) can be achieved at low frequencies (≤600 Hz) by adjusting the positions of individual disks and other structural parameters. Furthermore, a feedforward network-design approach based on machine learning models is proposed and experimentally demonstrated for the meta-absorber. Two types of meta-absorber features and different machine learning algorithms, including artificial neural network (ANN), k-Nearest Neighbor (kNN), and radial basis function neural network (RBFN), are utilized to compare their predictive abilities. The implementation of machine learning models allows for efficient design optimization of the meta-absorber's geometrical parameters to achieve specific functionalities without relying on physical models. The geometrical parameters are adaptively tuned during the design process to enhance sound absorption performance, particularly at low frequencies. The predicted sound absorption values are compared with the experimental results, and among the machine learning algorithms employed, ANN demonstrates the best performance, followed by the kNN method.

Modeling, Feedforward Control, and Constrained Trajectory Generation for a Concrete Conveyance System

Abstract Automated fabrication of concrete elements requires precise control of outgoing volume flow. Since measurements of the volume flow or the total extruded volume are challenging during production, a feedforward control is required, which in turn necessitates a model of the dynamics at hand. In this paper, a concrete conveyance system with two pumps is considered. These are mounted at the inlet and outlet of the hose, respectively. Based on previous work, a dynamical model with concentrated parameters is derived, which is motivated on the general distributed dynamics of two-phase flow. The model is validated by experiments, with parameters calculated by a tailored estimation approach. Based thereon, a feedforward control approach is presented and tested in simulation, exploiting the approximate flatness of the system. The flat parametrization is then used for constrained, smooth setpoint transition of the volume flow with minimal transition time.

Capacitor Voltage Weighted Full Feedforward Control Strategy for Adaptability Enhancing of LCL Grid-Connected Inverter Under Weak Grid

A capacitor-voltage full feedforward strategy is generally utilized to enhance the stabilization of the LCL grid-connected inverter. Meanwhile, the digitally controlled retardation leads to a phase lag of about 90° in the inverter’s export impedance on account of the introduction of the feed-forward loop, which influences the adaptation of the grid-connected inverter for a weakened grid. This investigation proposed a capacitor voltage-weighted full feedforward approach, which uses weighting coefficients to correct the inverter’s output impedance at various frequencies. In this way, the robustness of the grid-connected inverter operating at weak network conditions is improved. The proposal is validated by a combination of dramatic theoretical arguments and experimental findings.

Robust fixed-order [formula omitted] controller synthesis for multi-phase converters with input voltage feed-forward

Generated DC voltages of batteries, fuel cells, or supercapacitors as power sources connected in series with DC-DC converters change according to operating conditions. In such cases, the output voltage stability of the multi-phase DC-DC converters is an issue, particularly being subject to noises in the input voltage and load current disturbances. Input voltage feed-forward (IVFF) compensation approaches can improve the output voltage stability of the DC-DC converters in such scenarios. This paper proposes a robust fixed-order H∞ based IVFF control framework by considering sensor noise, input voltage, and load current disturbances to improve output voltage stability and reference tracking performance. The mismatched disturbance transfer functions depending on parasitic resistances are derived to include in the controller synthesis process. An iterative convex-concave optimization-based robust fixed-order controller design algorithm is proposed to optimize the feedback and feed-forward controllers under the stability and performance constraints. This algorithm allows the designing of the controllers simultaneously. The feed-forward controller, whose structure is independent of the system transfer functions, is optimized without any model inversion or extra design steps. The performance of the proposed controller is compared with an integral linear quadratic regulator controller, a sliding mode PID controller augmented with a nonlinear disturbance observer, and a fixed-order robust controller designed with non-smooth techniques. Real-time implementation of the fixed-order control systems is performed on a 2-leg 200 W interleaved boost converter. The performances are discussed under external disturbance effects to show the validity of the proposed control approach.

Iterative Learning Control for Compliant Underactuated Arms

Operations involving safe interactions in unstructured environments require robots with adapting behaviors. Compliant manipulators are a promising technology to achieve this goal. Despite that, some classical control problems such as following a trajectory are still open. A typical solution is to compensate the system dynamics with feedback loops. However, this solution increases the effective robot stiffness and jeopardizes the safety property provided by the compliant design. On the other hand, purely feedforward approaches can achieve good tracking performance while preserving the robot intrinsic compliance. However, a feedforward control framework for robots with passive elastic joints is still missing. This article presents an iterative learning control algorithm for purely feedforward trajectory tracking for compliant underactuated arms. Each arm is composed of active elastic joints and a generic number of passive ones connected through rigid links. We prove the convergence of the iterative method, also in the presence of uncertainties and bounded disturbances. Different output functions are analyzed providing conditions, based on the system inertial properties that ensure the algorithm applicability. Additionally, an automatic selection of the learning gain is proposed. Finally, we extensively validate the theoretical results with simulations and experiments.