The affine projection sign algorithm (APSA) has garnered significant attention in adaptive filtering due to its exceptional robustness and reduced computational demands. Nevertheless, the inherent use of a fixed projection order in APSA can compromise filtering accuracy and convergence speed. To address this issue, we introduce an innovative strategy for dynamically updating the projection order, resulting in an enhanced version of APSA called the evolving order based APSA (E-APSA). This evolving strategy compares the instantaneous power of output error to a threshold determined by the steady-state mean-square error of APSA, thereby enabling variable projection orders. Furthermore, we provide computational complexity and convergence analyses for E-APSA. Simulation results demonstrate that, compared to other related algorithms, E-APSA offers a significantly faster convergence rate while maintaining competitive steady-state misalignment.

In this paper, jointly with the exponential weighted generalized maximum correntropy (GMC) criterion and the linear constraint framework, we derive a recursive constrained adaptive filtering algorithm named recursive constrained GMC with forgetting factor (FF-RCGMC). In addition, due to a lack of constraint information during the learning process, FF-RCGMC will diverge or even fail to work after some iterations. Therefore, we propose a more stable version by introducing a constraint-forcing strategy into FF-RCGMC and call this robust type as constraint-forcing FF-RCGMC (CFFF-RCGMC). Some simulation results in system identification under non-Gaussian noisy environments validate the effectiveness of CFFF-RCGMC.

In robust adaptive filtering, the affine projection sign algorithm (APSA) is widely used in practice due to the desirable convergence behavior and low computational cost. Several variants of APSA have been derived from the variable step-size method and the combination strategy. For APSA, to further improve the filtering performance, this paper proposes a new filtering performance-enhanced (APSA) by optimizing a weighted cost function. Specifically, motivated by individual-weighting factors, our proposed algorithm, i.e., IWF-APSA, uses an individual-weighting factor to a corresponding input signal, while it has similar computational complexity to APSA. We also conduct the mean-square convergence analysis of IWF-APSA. The simulation results show that IWF-APSA achieves a better filtering performance in terms of convergence rate and filtering accuracy in system identification.

The extensive development of complex applications in embedded devices has driven the rapid development of edge computing, which provides powerful processing capabilities to the edge network. In this context of development, task offloading has received widespread attention as one of the key issues in edge computing. However, since a task usually consists of multiple subtasks with dependencies, the current computing subtask on the local or edge server must wait for the completion of the previous dependent subtask. As a result, existing offloading schemes are often constrained by the complexity of the task topology. Therefore, this paper is aimed at tasks with dependencies described as Directed Acyclic Graphs (DAGs) in devices. First, we optimize the order of subtasks with dynamic priorities. Meanwhile, we employ task duplication to reduce communication latency and thus overall task completion time. Additionally, to support the edge computing environment of multi-user and multi-server, game theory is used to find the optimal offload position for each user, so as to obtain the minimum average computing delay of all tasks. Experimental results show that the proposed algorithm outperforms existing algorithms in terms of task completion delay.

In the current landscape of industrial IoT time series databases, there is a lack of sufficient consideration for high-speed data writing and a failure to fully harness the temporal correlations inherent in the data. To address this issue, our study introduces OC-TS, a time-series data storage scheme based on Open-Channel Solid-State Disks (OCSSD).It is designed to eliminate capacity loss due to timestamps and the need for complex indexing when storing time-series data, facilitate fast retrieval of multiple sensors through time ranges. Simultaneously, we provide reserved space, allowing for such flexibility in accommodating new sensors or addressing sensor damage.

Existing multi-FPGA architectures often leverage high-speed interconnect technologies to achieve higher performance by exploiting ample communication bandwidth. In this paper, we propose an effective mapping approach for accelerating CNNs on bandwidth-constrained distributed multi-FPGA architectures. We formulate the system-level mapping problem and then introduce a method based on Genetic Algorithm (GA) and Mixed-Integer Nonlinear Programming (MINLP) to attain optimal solutions.

The recursive median filter has stronger noise at-tenuation capability than the median filter, especially for high-intensity and irregularly distributed noise. However, the recursive operation prevents recursive median filter from being pipelined, which leads to the recursive median filter being not real-time enough to be widely applied. This paper presents an FPGA implementation of low-latency recursive median filter. The proposed architecture completes the median calculation of the current window and the data pre-processing of the next window in one clock cycle, thereby reducing the calculation complexity of each median. The results show that for 5x5 window, the proposed recursive median filter core operates at a maximum frequency of 334 MHz on a zynq ultrascale+ FPGA device, which meets the real-time processing requirements for Full High Definition(FHD) images.

Unmanned aerial vehicle (UAV) working as an aerial station can gather the instantaneous information to guarantee the low-latency communication for the smart grid network. In this paper, we firstly construct a practical model of the end-to-end delay with considering the bit-error-ratio (BER) requirement of the communication link, and optimize the UAV’s trajectory to minimize the end-to-end delay between the UAV and the smart grid terminals, in which the critical-mission terminals (CMTs) or non-critical-mission terminals (NCMTs) send the individual information to the flying UAV. Although this non-convex problem is difficult to solve, we propose a trajectory design scheme based on Q-learning. To reduce the delay of CMTs, we design the different reward function for CMTs and NCMTs. The promising advantage of proposed scheme is that some NCMTs closed to CMTs may obtain the priority service to reduce the waiting delay. Simulation results show that our proposed scheme obtains almost 17% performance gain comparing to the benchmark schemes.