Search

Existing video compressive sensing (CS) techniques with fixed sampling rates can deliver satisfactory reconstructed quality but necessitate large transmission bandwidth. To overcome this challenge, region-of-interest (ROI)-based CS algorithms have been introduced to allocate different coding resources between ROI and non-ROI segments. However, neglecting non-ROI excessively in these algorithms leads to unsatisfactory average quality for the eventual reconstruction. In this article, we integrate the ideas of these methods and propose a novel adaptive video CS approach using a low-complexity ROI detection method in the compressed domain. The ROI is detected and sampled by calculating the measurement variance between the reference frame and the subsequent frames. Conversely, the non-ROI is not transmitted but will be reconstructed by utilizing the reference frame through the corresponding position information. In addition, we present a compact method for adapting the threshold value, which allows each frame of a video to have a unique threshold rather than an artificially predetermined fixed value. Moreover, a reference-frame-updating strategy is developed to improve the versatility of the entire framework. Compared to state-of-the-art counterparts, extensive experimental results have demonstrated that our proposed methods achieve superior performance while tackling diverse scenes and using a lower sampling rate.

The duck curve is becoming a global problem in energy technology due to the rapid increase in solar power adoption and the rise of prosumers. To address this issue, a resource aggregator (RA) has emerged to provide flexible solutions through aggregating the prosumers and demand response such as dynamic pricing. This paper proposes an optimal strategy for the RA that dispatches dynamic pricing to the prosumers and leverages the battery system at both RA and prosumer levels. The proposed method is based on a model-free deep reinforcement learning (DRL) algorithm to optimize each prosumer’s retail prices and schedule usage of the RA’s battery power station. An objective reward function is used to maximize the RA’s profit, minimize the prosumer’s cost, and maximize the improvement of the duck curve. The performance of the proposed DRL-based strategy was demonstrated by simulation experiments using actual wholesale price, demand, and PV generation data. The results show that the proposed strategy can improve the standard deviation and peak-to-average ratio of net load by up to 57.1% and 23%, respectively.

Multicore task scheduling that determines the execution order of tasks and mapping on multiple cores is one of the most crucial design processes in order to fully exploit the potential of multicore systems. This paper studies a task scheduling problem for mouldable fork-join tasks with inter and intra-task communications. Unlike existing works, our work assumes that a task can be split into multiple threads, and the threads can run independently in a data-parallel fork-join manner. Our proposed techniques decide the number of cores and the execution start time for each task on the assumption of inter and intra-task communications at the same time during scheduling. The proposed techniques are based on integer programming formulation. The goal is to minimise the overall schedule length. We evaluate the performance of our proposed techniques and the effect of the communications on scheduling results.

High‐level synthesis (HLS), which automatically synthesizes a register‐transfer level (RTL) circuit from a behavioral description written in a high‐level programming language such as C/C++, is becoming a more popular technique for improving design productivity. In general, HLS tools often generate a circuit with a larger area than those of hand‐designed ones. One reason for this issue is that HLS tools often generate multiple instances of the same module from a function. To eliminate such a redundancy in circuit area in HLS, HLS tools are capable of sharing modules. Function‐level module sharing at a behavioral description written in a high‐level programming language may promote function reuse to increase effectiveness and reduce circuit area. In this paper, we present two HLS techniques for module sharing at the function level.

This paper studies a task scheduling problem which schedules a set of data-parallel tasks on multiple cores. Unlike most of previous literature where each task is assumed to run on a single core, this work allows individual tasks to run on multiple cores in a data-parallel fashion. Since the scheduling problem is NP-hard, a couple of heuristic algorithms which find near-optimal schedules in a short time were proposed so far. In some cases, however, exactly-optimal schedules are desired, for example, in order to evaluate heuristic algorithms. This paper proposes an exact algorithm to find optimal schedules. The proposed algorithm is based on depth-first branch-and-bound search. In our experiments with up to 100 tasks, the proposed algorithm could successfully find optimal schedules for 135 test cases out of 160 within 12 hours. Even in cases where optimal schedules were not found within 12 hours, our algorithm found better schedules than state-of-the-art heuristic algorithms.

This paper proposes a task scheduling algorithm for multi/many-core systems. To increase the quality of results on the low computational complexity, our algorithm uses two static priorities switched during task scheduling. In our experiments, we compared the proposed algorithm with a state-of-the-art algorithm. The experimental results show that the proposed algorithm yields shorter scheduling length than our previous research.

The DC microgrid system is one of hopeful candidate of the community-level power systems including renewable energy sources. This paper proposes an abstract system model of DC microgrid systems for fast and accurate simulation toward long-term virtual experiment of the DC microgrid system. We constructed the abstract system model based on the specification of the DC microgrid system. Experimental results show that the simulation results are valid comparing with the real measured data.

Under dynamic pricing, stable and accurate electricity price forecasting on the demand side is essential for efficient energy management. We have developed a new electricity price forecasting model that provides consistently accurate forecasts. The base prediction model decomposes the time series using wavelet transform and then predicts it by Long Short-Term Memory. Previous studies using this model have always decomposed time series in the same way without changing the mother wavelet. However, this makes it difficult to respond to changes in time series that vary daily or seasonally. Therefore, we periodically switch the mother wavelet, i.e., flexibly change the time series decomposition method, to achieve stable and highly accurate electricity price forecasting. In an experiment, the model improved prediction accuracy by up to 42.8% compared to prediction with a fixed mother wavelet. Experimental results show that the proposed flexible forecasting method can consistently provide highly accurate forecasts.

Function-level profiling is crucial for optimised embedded software which has severe constraints on hardware resources, low power consumption, real-time responses, and so on. In this work, we provide a fast and reliable solution by utilising an instruction set simulator named QEMU and creating an analyser tool. We developed a tracing module inside the simulator to trace execution information of software at run-time and record it in a log file. Our implementation takes advantage of the dynamic binary translation of QEMU to keep its speed fast and use an analyser tool to analyse the log file and create a function-level profile. We implemented our solution methodology for the ARM architecture, and evaluated it with various kinds of embedded applications.