We present Cacheback Decoding, a trainingfree and model-agnostic speculative decoding method that exploits the locality in language to accelerate Large Language Model (LLM) inference.Cacheback leverages only Least Recently Used (LRU) cache tables of token n-grams to generate draft sequences.Cacheback achieves state-of-the-art performance among comparable methods despite its minimalist design, and its simplicity allows easy integration into existing systems.Cacheback also shows potential for fast adaptation to new domains.

Microcontroller-based embedded systems are vulnerable to memory safety errors and must be robust and responsive because they are often used in unmanned and mission-critical scenarios. The Rust programming language offers an appealing compile-time solution for memory safety but leaves stack overflows unresolved and foils zero-latency interrupt handling. We present Hopter, a Rust-based embedded operating system (OS) that provides memory safety, system robustness, and interrupt responsiveness to embedded systems while requiring minimal application cooperation. Hopter executes Rust code under a novel finite-stack semantics that converts stack overflows into Rust panics, enabling recovery from fatal errors through stack unwinding and restart. Hopter also employs a novel mechanism called soft-locks so that the OS never disables interrupts. We compare Hopter with other well-known embedded OSes using controlled workloads and report our experience using Hopter to develop a flight control system for a miniature drone and a gateway system for Internet of Things (IoT). We demonstrate that Hopter is well-suited for resource-constrained microcontrollers and supports error recovery for real-time workloads.

Emerging large language model (LLM) applications involve diverse reasoning strategies and agentic workflows, straining the capabilities of existing serving systems built on a monolithic token generation loop. This paper introduces Pie, a programmable LLM serving system designed for flexibility and efficiency. Pie decomposes the traditional generation loop into fine-grained service handlers exposed via an API and delegates control of the generation process to user-provided programs, called inferlets. This enables applications to implement new KV cache strategies, bespoke generation logic, and seamlessly integrate computation and I/O—entirely within the application, without requiring modifications to the serving system. Pie executes inferlets using WebAssembly, benefiting from its lightweight sandboxing. Our evaluation shows Pie matches state-of-the-art performance on standard tasks (3-12% latency overhead) while significantly improving latency and throughput (1.3×-3.4× higher) on agentic workflows by enabling application-specific optimizations.

An adaptive fuzzy command filtering backstepping technique is given for strict-feedback fractional-order uncertain nonlinear systems. The controlled system includes unknown nonlinear functions, as well as state and input quantization. Considering fractional-order nonlinear systems that do not satisfy matching conditions, unknown nonlinear functions are approximated by fuzzy logic systems, and a sector bounded quantizer is used to quantify all input and state variables. During the plan process, command filtering backstepping scheme is used to avoid the use of nonsmooth states. Subsequently, in order to ensure the boundedness of a series of errors caused by continuous original states for stability analysis and discontinuous quantization states for control, a sufficiently smooth fractional-order projection operator is proposed. In addition, the fractional-order uniformly bounded criterion has been established and strictly proven, which solves the problem of uniformly bounded error signals in the fractional-order sense under the premise of known parameter boundedness. Thus, the boundedness of all closed-loop signals is ensured by the fractional-order uniformly bounded criterion. Finally, the simulation results have confirmed the efficacy of the method.

This paper presents a nonlinear filters-based adaptive fuzzy control design for strict-feedback nonlinear systems with unknown asymmetric dead-zone output and virtual control coefficients. First, a novel smooth approximation of the non-smooth asymmetric output dead-zone nonlinearity is constructed to improve the approximation performance and speed. Then, to dilute the effects of unknown dead-zone and virtual control coefficients, an adaptive compensation mechanism is presented by utilizing projection operators techniques and the properties of fuzzy basis functions and hyperbolic tangent functions. Further, a novel adaptive backstepping control design method based on the nonlinear filters is proposed, which not only avoids the issue of explosion of complexity inherent in the backstepping procedure, but also completely compensates the effects of the boundary errors generated by the introduced filters in spite of unknown virtual control coefficients. From the properties of the smooth projection operator, and based on Lyapunov synthesis, it is shown that all closed-loop signals are ensured uniformly bounded. Finally, the applicability of the proposed control method is rigorously verified by a example of one-link manipulator with a brushed dc motor.Note to Practitioners—Strict-feedback nonlinear systems can be utilized to model many practical engineering systems, such as one-link manipulator, permanent magnet synchronous motor, robots with flexible joints, spacecrafts, etc. Due to the physical limitations and other factors, non-smooth dead-zone nonlinearity often presents in system components, which will influence the control performance and threat the operation security. Therefore, this paper considers the control design problem for strict-feedback nonlinear systems with unknown asymmetric dead-zone output nonlinearity, and the proposed control algorithm is successfully applied to one-link manipulator with a brushed dc motor. In addition, the virtual control coefficients are often assumed to be known constants or known nonlinear functions with known bounds in the existing results, which is unrealistic on account of the economic costs and environmental factors. To erase such limitation, this paper presents a novel adaptive compensation mechanism based on smooth auxiliary function dependent on the property of tanh-function and the projection operator technique.

Stack unwinding is a well-established approach for handling panics in Rust programs. However, its feasibility on resource-constrained embedded systems has been unclear due to the associated overhead and complexity. This paper presents our experience of implementing stack unwinding and panic recovery within a Rust-based soft real-time embedded operating system. We describe several novel optimizations that help achieve adequate performance for a flying drone with a CPU overhead of 2.6% and a storage overhead of 26.0% to recover from panics in application tasks and interrupt handlers.

This article presents an adaptive fuzzy output-feedback dynamic surface control (DSC) for the nonlinear systems with dead-zone output nonlinearity. First, a novel smooth approximation of the output dead zone is given to conveniently fuse with the backstepping technique. After that a nonlinear fuzzy state observer is constructed to estimate the unmeasurable states, by employing the fuzzy logic systems for identifying the unknown compounded nonlinear functions, which releases the limitation in the existing references that the state observer needs to be linear in the presence of dead-zone output nonlinearity. Then, to reduce the effect of unknown dead-zone coefficients, an adaptive compensation mechanism is introduced by using the characteristic of hyperbolic tangent function, which can replace the widely used Nussbaum-type function-based control strategy. Furthermore, a novel DSC method based on the nonlinear filters is proposed, which not only avoids the issue of explosion of complexity inherent in the backstepping procedure, but also improves the system control performance. Under the certain assumptions, the stability of the closed-loop system is proved by use of Lyapunov function stability theory. Finally, the applicability of the proposed control method is rigorously verified by a single-link robot arm simulation example.

Continuous mobile vision is becoming increasingly important as it finds compelling applications which substantially improve our everyday life. However, meeting the requirements of quality of experience (QoE) diversity, energy efficiency and multi-tenancy simultaneously represents a significant challenge. In this paper, we present SCYLLA, an FPGA-based framework that enables QoE-aware continuous mobile vision with dynamic reconfiguration to effectively address this challenge. SCYLLA pre-generates a pool of FPGA design and DNN models, and dynamically applies the optimal software-hardware configuration to achieve the maximum overall performance on QoE for concurrent tasks. We implement SCYLLA on state-of-the-art FPGA platform and evaluate SCYLLA using drone-based traffic surveillance application on three datasets. Our evaluation shows that SCYLLA provides much better design flexibility and achieves superior QoE trade-offs than status-quo CPU-based solution that existing continuous mobile vision applications are built upon.

High-speed rail (HSR) systems potentially provide a more efficient way of door-to-door transportation than airplane. However, they also pose unprecedented challenges in delivering seamless Internet service for on-board passengers. In this paper, we conduct the first large-scale disconnection-centric measurement study of TCP performance over LTE on HSR. Our measurement targets the main HSR route in China operating at 300/350 km/h. We performed extensive data collection obtaining 378.3 GB data collected over 56639 km of trips. Leveraging such a unique dataset, we measure important performance metrics such as TCP goodput, latency and loss rate across different congestion control algorithm, mobile carrier, and different train speed. We further develop the LTE disconnection taxonomy, and conduct a in-depth correlation study between TCP stall and LTE disconnection. Our findings reveal the networking performance on today's HSR environment “in the wild”, as well as identify several root causes of performance inefficiencies, which together highlight the need to develop dedicated protocol mechanisms that are friendly to extreme mobility.

This paper presents a novel adaptive fuzzy backstepping dynamic surface control (DSC) scheme for a class of single-input single-output strict-feedback fractional-order uncertain nonlinear systems. The controlled systems contain unknown nonlinear functions and unknown external disturbances. Fuzzy logic systems are employed for approximating the unknown nonlinear functions. Further, an auxiliary function is introduced into the control function to simultaneously compensate the unknown external disturbance and the approximation error caused by fuzzy approximation, which erases the possible chattering phenomenon in the existing results. Meanwhile, a new DSC method based on the fractional-order filter is proposed to avoid the issue of explosion of complexity inherent in the backstepping procedure, which releases the limitation that the fractional-order derivative of the intermediate control function needs to be completly known in the existing references. Under certain assumptions, the stability of the closed-loop system is proved by using the fractional-order Lyapunov function stability criterion. Finally, contrastive simulation results are provided to validate the effectiveness of our proposed control strategy.

This article presents two adaptive fuzzy output-feedback dynamic surface control schemes for single-input-single-output (SISO) strict-feedback fractional-order uncertain nonlinear systems under the complete smooth nonlinearity and partial piecewise-smooth nonlinearity, respectively. For the former, complete smooth nonlinear functions are approximated by employing fuzzy logic systems, based on which a novel fractional-order reduced-order observer is constructed to estimate the unmeasurable states. For the latter, a new switched approximation and estimate strategy is proposed to deal with the problem of piecewise-smooth nonlinearity. Meanwhile, fuzzy basis function property is utilized to eliminate the assumption requirement that the nonlinear functions must satisfy the Lipschitz condition inherent in the output-feedback control, which is suitable for two cases. Under certain assumptions, the stability of two closed-loop systems are proved via fractional-order Lyapunov function stability criterion. Specifically, the stability proof of the second case is built on the foundation of the common Lyapunov function method. Finally, two diverse Chua's circuit simulation examples are provided to, respectively, validate the effectiveness of the proposed two control strategy.

The problem of adaptive decentralized fault-tolerant control is investigated in this paper for the pure-feedback nonlinear switched large-scale systems. The considered systems possess the unknown compounded nonlinearities (unknown nonlinear functions, unknown interconnected terms, unknown nonlinear function faults, and unknown actuator dead-zone nonlinearity). The radial basis function neural networks are utilized for identifying the unknown compounded nonlinear functions so that the problems of unknown nonlinearities can be solved. A switched neural networks k -filter observer is constructed to estimate the unmeasurable states, deal with the unknown actuator dead-zone nonlinearity and compensate the nonlinear function faults simultaneously. Then, by combining the average dwell time theory, adaptive decentralized control method, and under the framework of backstepping design, an adaptive output-feedback fault compensation control method is presented. Under the certain assumptions, the stability of the closed-loop system is proved by use of Lyapunov function stability theory. Finally, the applicability of the proposed controller is well carried out by a double-inverted-pendulums simulation example.

In this paper, an adaptive fuzzy fault-tolerant control approach is proposed for a class of nontriangular structure nonlinear systems, which contain immeasurable states and unknown actuator faults (actuator loss-of-effectiveness and bias fault). It should be noted that if the existing approaches are employed for nontriangular structure nonlinear systems, the algebraic loop problem may occur. In this study, fuzzy logic systems are employed to approximate unknown nonlinear functions, and a fuzzy state observer is designed to estimate immeasurable states. Then, based on the property of performance function and simple barrier Lyapunov function design method, the prescribed output tracking error dynamic performance and the corresponding stability are guaranteed. By using the parameter estimation technique, a new fault compensation strategy is developed to relax the requirement that the efficiency indicator must be known. The stability of the closed-loop system is proved by using the Lyapunov function stability theory. Finally, a simulation example is given to validate the effectiveness of the proposed control strategy.

The problem of adaptive fuzzy output-constrained tracking fault-tolerant control (FTC) is investigated for the large-scale stochastic nonlinear systems of pure-feedback form. The nonlinear systems considered in this paper possess the unstructured uncertainties, unknown interconnected terms and unknown nonaffine nonlinear faults. The fuzzy logic systems are employed to identify the unknown lumped nonlinear functions so that the problems of structured uncertainties can be solved. An adaptive fuzzy state observer is designed to solve the nonmeasurable state problem. By combining the barrier Lyapunov function theory, adaptive decentralized and stochastic control principles, a novel fuzzy adaptive output-constrained FTC approach is constructed. All the signals in the closed-loop system are proved to be bounded in probability and the system outputs are constrained in a given compact set. Finally, the applicability of the proposed controller is well carried out by a simulation example.

In cognitive radio networks, knowledge of the position of the primary users is very important as it can be used to avoid harmful interference to the primary users, while at the same time be exploited to improve the spectrum utilization. In this paper, a semi range-based localization algorithm is proposed for the secondary users in cognitive radio networks to estimate the positions of the primary users. The basic idea of the proposed algorithm is to take advantage of the estimated detection probabilities, which can be obtained from the binary detection indictors of the secondary users, in order to estimate the distances between themselves and the primary users. The accuracy of the proposed localization algorithm is further improved by introducing an iterative least squares algorithm. The Cramer-Rao lower bound of the mean square error of the proposed localization estimator is also derived. Extensive simulations will then show that the actual mean square error achieved by the proposed localization algorithm is reasonably close to the lower bound, which demonstrates that the proposed method is near optimal.

Cognitive Radios allow secondary users to share the spectrum with primary users. The fairness between secondary users in Cognitive Radio networks is an important issue. This paper analyzes the fairness based on the distributed opportunistic spectrum access scheme. A fair multiple access scheme using fast catch-up strategy is proposed. The theoretical result of the first passage time to achieve fairness is given. Simulation results demonstrate that the Fair OSA scheme can achieve fairness much faster when a new secondary user accesses the spectrum. A complex practical scenario is simulated to validate the fairness in more general situations.