Although federated learning stands as a prominent distributed learning paradigm across multiple clients, it has grappled with a dilemma: the choice between developing a singular global model to promote global generalization or nurturing personalized models to accommodate personalization. In this article, a federated learning with warm-start (WarmFed) is proposed to obtain both robust global and personalized models. First, a client knowledge-driven initialization approach named warm-start is introduced, which provides resilient global information to assist subsequent training. Then, a dynamic self-distillation (DSD) strategy is proposed to obtain more resilient personalized models and a model fine-tuning (MFT) mechanism based on synthetic data is designed to improve performance of the global model on the server side. The collaborative deployment of DSD and MFT exploits potential to achieve both generalization and personalization under warm-start. Comprehensive experiments underscore the superiority of our approach across both global and personalized models under domain shift and label skew.

As is known, the problem of making a trained deep reinforcement learning (DRL) model to adapt to a new environment different from where it trains remains open. In this article, DRL-based techniques are investigated and deployed for the robotic motion planning task in locally changed environments. First, a dynamic-entropy actor-critic is proposed to automatically adjust the coefficient of the entropy cluster to enable the agent to efficiently learn the potentially optimal policy. Second, a tutoring-guiding mechanism is established to heuristically explore and sufficiently exploit valuable experience. Third, a mechanism consisting of progressive planner and inverse kinematics mapper is designed to cope with localized changes in the environment, which enables the robot manipulator to rapidly adapt to the newly changed environment. Finally, experimental results have verified the superior performance of the proposed approaches.

Deep hashing retrieval has gained widespread use in big data retrieval due to its robust feature extraction and efficient hashing process. However, training advanced deep hashing models has become more expensive due to complex optimizations and large datasets. Coreset selection and Dataset Condensation lower overall training costs by reducing the volume of training data without significantly compromising model accuracy for classification task. In this paper, we explore the effect of mainstream dataset condensation methods for deep hashing retrieval and propose IEM (Information-intensive feature-Embedding Matching), which is centered on distribution matching and incorporates model and data augmentation techniques to further enhance the feature of hashing space. Extensive experiments demonstrate the superior performance and efficiency of our approach.

Although robotic trajectory generation problem has been extensively investigated, existing solutions are almost customized to specific robot geometry, and generalized schemes are yet to be explored. In this article, a general motion planning framework based on deep reinforcement learning (DRL) and artificial neural networks (ANNs) is proposed for robot with arbitrary geometry. First, a unique screening and grafting mechanism is established to improve the policy learning by exploiting valuable experience sufficiently. Second, based on the reward-oriented characteristics of DRL, a forward progression mechanism is proposed to facilitate the path planning for complicated tasks. Third, a structure consisting of an adventurer and conservator algorithm with automatic optimization and an ANN-based mapper is designed integrally to derive the inverse kinematics solutions without considering the robot geometry. Finally, experimental results have verified the superior performance of the proposed approach.Note to Practitioners—This article is aiming to provide a general method to solve the problem of motion planning for robots via deep reinforcement learning (DRL) and artificial neural networks (ANNs). Compared to the existing approaches, which are highly specialized and limited to robots with specific geometries, and often cumbersome, our method can be easily applied to robots with arbitrary geometries and has good generalization, where to simplify the training based on DRL for diverse practical motion planning tasks, a universal forward progression mechanism is used to partition a complex task into multiple successive simple phases. Furthermore, an ANN-based mapper can cope with the burdensome inverse kinematics of robots with both common and uncommon geometries, especially the ones with redundant degrees of freedom. The proposed method is also validated to be superior to other state-of-the-art methods in real-world experimental studies.

Motion planning and its optimization is vital and difficult for redundant robot manipulator in an environment with obstacles. In this article, a general motion planning framework that integrates deep reinforcement learning (DRL) is proposed to explore the length-optimal path in Cartesian space and to derive the energy-optimal solution to inverse kinematics. First, based on the maximum entropy framework and Tsallis entropy, a DRL algorithm with clipped automatic entropy adjustment is proposed to make the agent to be qualified to cope with diverse tasks. Second, a path planning structure that combines traditional path planner and DRL is proposed, which integrates the powerful exploration capability of the former and exploitation of experience replay of the latter to enhance the planning performance. Third, based on the exploration ability of DRL and the nonlinear fitting ability of artificial neural networks, a structure is proposed to provide an energy-optimal inverse kinematics solution for redundant robot manipulators. Finally, experimental results on both simulated and real-world customized scenarios have verified the performance of the proposed work.

Power control and optimization are both crucial for the proper operation of a microgrid. However, in existing research, they are usually studied separately. Active and reactive powers are either maintained to constant values at device level or optimized at system level without considering frequency and voltage control of distributed converters. In this article, a holistic power control and optimization strategy is proposed for microgrids. Specifically, a model predictive control incorporated with the droop method is developed at device level to achieve load sharing and flexible power dispatching among distributed energy resources, which is feasible for both islanded and grid-connected modes. In addition, an evolutionary particle swarm optimization algorithm is designed at system level to generate the optimal active and reactive power setpoints, which are then sent to the device level for controlling inverters. The proposed power optimization scheme is able to mitigate voltage deviations and minimize the operational cost of the microgrid. Comprehensive case studies and real-time simulator test are provided to demonstrate the feasibility and efficacy of the proposed power control and optimization strategy.

Trajectory generation for redundant manipulators based on inverse kinematics (IK) still faces some restraints, as it lacks universal IK calculation or specific trajectory generation methods that are suitable for robots with arbitrary degrees of freedom. In this article, the IK-free trajectory generation for robot manipulators is formulated as a Markov decision process and implemented by a general method based on deep reinforcement learning. First, an extensively explored and evaluated actor-critic (E3AC) algorithm that can make diverse action explorations and comprehensive evaluations is designed to solve the trajectory generation problem. Second, a dual-memory structure with expert-guided policy learning strategy is proposed to further enhance the performance of the algorithm in the early training period by additional successful experiences and performing an increasingly unbiased data sampling. Third, a fuzzy feedback reward mechanism that can directly establish a mapping from the abundant state variables to the self-tuning reward is constructed, instead of puzzling out an explicit function to feature the complex relations among the control objects. Finally, the comparative experimental results show that, the proposed approach is more efficient in algorithm convergence and reward calculation, and is more qualified for complex tasks with strong randomness.

This paper concerns the problem of output feedback tracking (OFT) control with bounded torque inputs of robot manipulators, and proposes a novel saturated OFT controller based on fuzzy self-tuning proportional and derivative (PD) gains. First, aiming to accomplish the whole closed-loop control with only position measurements, a linear filter is involved to generate a pseudo velocity error signal. Second, different from previous strategies, the arctangent function with error-gain is applied to ensure the boundedness of the torque control input, and an explicit system stability proof is made by using the theory of singularly perturbed systems. Moreover, a fuzzy self-tuning PD regulator, which guarantees the continuous stability of the overall closed-loop system, is added to obtain an adaptive performance in tackling the disturbances during tracking control. Simulation showed that the proposed controller gains more satisfactory tracking results than the others, with a better dynamic response performance and stronger anti-disturbance capability.