Removing Feasibility Conditions on Adaptive Neural Tracking Control of Nonlinear Time-Delay Systems With Time-Varying Powers, Input, and Full-State Constraints

Sample-efficient and occlusion-robust reinforcement learning for robotic manipulation via multimodal fusion dualization and representation normalization

Recent advances in visual reinforcement learning (visual RL), which learns from high-dimensional image observations, have narrowed the gap between state-based and image-based training. However, visual RL continues to face significant challenges in robotic manipulation tasks involving occlusions, such as lifting obscured objects. Although high-resolution tactile sensors have shown promise in addressing these occlusion issues through visuotactile manipulation, their high cost and complexity limit widespread adoption. In this paper, we propose a novel RL approach that introduces multimodal fusion dualization and representation normalization to enhance sample efficiency and robustness in robotic manipulation tasks involving occlusions - without relying on tactile feedback. Our multimodal fusion dualization technique separates the fusion process into two distinct modules, each optimized individually for the actor and the critic, resulting in tailored representations for each network. Additionally, representation normalization techniques, including LayerNorm and SimplexNorm, are incorporated into the representation learning process to stabilize training and prevent issues such as gradient explosion. We demonstrate that our method not only effectively tackles challenging robotic manipulation tasks involving occlusions but also outperforms state-of-the-art visual RL and state-based RL methods in both sample efficiency and task performance. Notably, this is achieved without relying on tactile sensors or prior knowledge, such as predefined low-dimensional coordinate states or pre-trained representations, making our approach both cost-effective and scalable for real-world robotic applications.

Fixed-Time Composite Neural Learning Control of Flexible Telerobotic Systems

This article is devoted to the fixed-time synchronous control for a class of uncertain flexible telerobotic systems. The presence of unknown joint flexible coupling, time-varying system uncertainties, and external disturbances makes the system different from those in the related works. First, the lumped system dynamics uncertainties and external disturbances are estimated successfully by designing a new composite adaptive neural networks (CANNs) learning law skillfully. Moreover, the fast-transient, satisfactory robustness, and high-precision position/force synchronization are also realized by design of fixed-time impedance control strategies. Furthermore, the "complexity explosion" issue triggered by traditional backstepping technology is averted efficiently via a novel fixed-time command filter and filter compensation signals. And then, sufficient conditions of system controller parameters and fixed-time stability are theoretically given by establishing the Lyapunov stability theorem. Besides, the absolute stability of the two-port networked system under complex transmission time delays is rigorously proved. Finally, simulations are performed with 2-link flexible telerobotic systems under two cases, results are presented to realistically verify the proposed control algorithm available.

Dynamic Gain Reduced-Order Observer-Based Global Adaptive Neural-Network Tracking Control for Nonlinear Time-Delay Systems

In this article, a globally adaptive neural-network tracking control strategy based on the dynamic gain observer is proposed for a class of uncertain output-feedback systems with unknown time-varying delays. A reduced-order observer with novel dynamic gain is proposed. An n th-order continuously differentiable switching function is constructed to achieve the continuous switching control of the system, thus further ensuring that all the closed-loop signals are globally uniformly ultimately bounded (GUUB). It is proved that by adjusting the designed parameters, the tracking error converges to a region which can be adjusted to be small enough. The effectiveness of the control scheme is demonstrated by two simulation examples.

Event-triggered tracking control of uncertain p-normal nonlinear systems with full-state constraints

This paper investigates the tracking control problem of uncertain p-normal nonlinear systems with full-state constraints via event-triggered mechanism. By skillful constructing an adaptive dynamic gain and a time-varying event-triggered strategy, a state-feedback controller is proposed to achieve practical tracking. The adaptive dynamic gain is incorporated to deal with the system uncertainties and eliminate the bad effect of the sampling error. A rigorous Lyapunov stability analysis method is put forward to verify that all the closed-loop signals are uniformly bounded and the tracking error converges into a prescribed arbitrary accuracy, and full-state constraints are not violated. Compared with the existing event-triggered strategies, the proposed time-varying event-triggered strategy is low-complexity without designing the hyperbolic tangent function.