Robust performance modelling and scheduling of distributed real-time systems

Distributed real-time (DRT) systems are increasingly requiring object profiling, scheduling and migration algorithms to respond to unpredictable transient changes in load and availability of resources in an open environment. A key challenge in developing distributed real-time systems is to meeting critical performance constraints when DRT systems become more difficult to predict in terms of the needs of the application, particularly those needs that are likely to change dynamically during execution of the application. The system design must combine temporal requirements and considerable uncertainty, together with the ability to swiftly adapt to changing environmental conditions. This paper proposes a robust DRT model that does not require precise system parameters. A multivariable H∞ controller is developed and provides utilisation and deadline guarantees in an unpredictable environment.     

Flexible hard real-time scheduling for deliberative AI systems

Constructing deliberative real-time AI systems is challenging due to the high execution-time variance in AI algorithms and the requirement of worst-case bounds for hard real-time guarantees, often resulting in poor use of system resources. Using a motivating case study, the general problem of resource usage maximization is addressed. We approach the issues by employing a hybrid task model for anytime algorithms, which is supported by recent advances in fixed priority scheduling for imprecise computation. In particular, with a novel scheduling scheme based on Dual Priority Scheduling, hard tasks are guaranteed by schedulability analysis and scheduled in favor of optional and anytime components which are executed whenever possible for enhancing system utility. Simulation studies show satisfactory performance on the case study with the application of the scheduling scheme. We also suggest how aperiodic tasks can be scheduled effectively within the framework and how tasks can be prioritized based on their utilities by an efficient algorithm. These works form a comprehensive package of scheduling model, analysis, and algorithms based on fixed priority scheduling, providing a versatile platform where real-time AI applications can be suitably facilitated.

The partitioned scheduling of sporadic task systems on multiprocessors

Abstract The DBF ^* algorithm of sporadic task systems on multiprocessors uses the approximation of the exact demand bound function on uniprocessor as a criterion. The systems which are feasible under the partitioned paradigm are flagged as “infeasible” sometimes. In this paper, we present a novel efficient DBF(eDBF) partitioned scheduling algorithm. A criterion which tracks the demand bound function exactly as needed is used to avoid the incorrect judgment in determining whether a processor can accommodate an additional task in the new algorithm. We give the pseudo code of the new algorithm on least-number processors and fixed-number processors respectively. Then, we prove the correctness of, and evaluated the effectiveness of this new algorithm. The experimental results demonstrate that eDBF has better performance than DBF ^* and Density algorithms.     

Decomposition based hybrid VNS–TS algorithm for distributed parallel factories scheduling with virtual corporation

The main purpose of this paper is to develop a decomposition based hybrid variable neighborhood search/tabu search (DVT) algorithm for multi-factory production network scheduling problem where a number of different individual factories collaborate despite their different objectives. It is assumed some of the network's factories are interested in total processing cost minimization whereas the remaining factories are interested in the production profits maximization. It is also assumed that jobs can migrate from their original factory to other factories but a transportation time is incurred. Our proposed algorithm comprises of a tabu search and a variable neighborhood search with several local search algorithms. In this hybridization, to improve the search ability of the algorithm, we make use of guiding principles with ordering of neighborhood structures by mixed integer linear programming relaxation. In the proposed algorithm, the parallel search strategy is designed for a scalar bi-objective. Multiple objectives are combined with L₁-metric technique then each sub-search procedure evolves separately until a good approximation of the Pareto-front is obtained. The non-dominated sets obtained from our algorithm and original heuristic (algorithm without ordering concept) are compared using three different indices. Furthermore, the problem is modeled as a mixed integer linear programming and solved by improved ϵ-constraint approach (IEA) with CPLEX solver. The results of comparisons between IEA and DVT algorithm showed the proposed algorithm yielded most of the solutions in the net non-dominated front.     

T–L plane-based real-time scheduling for homogeneous multiprocessors

We consider optimal real-time scheduling of periodic tasks on multiprocessors—i.e., satisfying all task deadlines, when the total utilization demand does not exceed the utilization capacity of the processors. We introduce a novel abstraction for reasoning about task execution behavior on multiprocessors, called T–L plane and present T–L plane-based real-time scheduling algorithms. We show that scheduling for multiprocessors can be viewed as scheduling on repeatedly occurring T–L planes, and feasibly scheduling on a single T–L plane results in an optimal schedule. Within a single T–L plane, we analytically show a sufficient condition to provide a feasible schedule. Based on these, we provide two examples of T–L plane-based real-time scheduling algorithms, including non-work-conserving and work-conserving approaches. Further, we establish that the algorithms have bounded overhead. Our simulation results validate our analysis of the algorithm overhead. In addition, we experimentally show that our approaches have a reduced number of task migrations among processors when compared with a previous algorithm.     

Stochastic analysis of real-time systems under preemptive priority-driven scheduling

Exact stochastic analysis of most real-time systems under preemptive priority-driven scheduling is unaffordable in current practice. Even assuming a periodic and independent task model, the exact calculation of the response time distribution of tasks is not possible except for simple task sets. Furthermore, in practice, tasks introduce complexities such as release jitter, blocking in shared resources, etc., which cannot be handled by the periodic independent task set model. In order to solve these problems, exact analysis must be abandoned for an approximated analysis. However, in the real-time field, approximations must not be optimistic, i.e. the deadline miss ratios predicted by the approximated analysis must be greater than or equal to the exact ones. In order to achieve this goal, the concept of pessimism needs to be mathematically defined in the stochastic context, and the pessimistic properties of the analysis carefully derived. This paper provides a mathematical framework for reasoning about stochastic pessimism, and obtaining mathematical properties of the analysis and its approximations. This framework allows us to prove the safety of several proposed approximations and extensions. We analyze and solve some practical problems in the implementation of the stochastic analysis, such as the problem of the finite precision arithmetic or the truncation of the probability functions. In addition, we extend the basic model in several ways, such as the inclusion of shared resources, release jitter or non-preemptive sections.

Iterative learning control for hyperbolic distributed parameter systems based on sensor–actuator networks

This paper proposes two kinds of iterative learning control (ILC) schemes for a class of the distributed parameter systems based on sensor–actuator networks which can be described by hyperbolic partial differential equations. A D-type ILC algorithm is first considered and the convergent condition of the output error is obtained via the contraction mapping methodology. Then, the PD-type ILC algorithm is considered in this hyperbolic distributed parameter systems based on sensor–actuator networks. Finally, a cable equation with air and structural damping is given to illustrate the effectiveness of the proposed methods.     

A software-based dynamic-warp scheduling approach for load-balancing the Viola–Jones face detection algorithm on GPUs

Face detection is a key component in applications such as security surveillance and human–computer interaction systems, and real-time recognition is essential in many scenarios. The Viola–Jones algorithm is an attractive means of meeting the real time requirement, and has been widely implemented on custom hardware, FPGAs and GPUs. We demonstrate a GPU implementation that achieves competitive performance, but with low development costs. Our solution treats the irregularity inherent to the algorithm using a novel dynamic warp scheduling approach that eliminates thread divergence. This new scheme also employs a thread pool mechanism, which significantly alleviates the cost of creating, switching, and terminating threads. Compared to static thread scheduling, our dynamic warp scheduling approach reduces the execution time by a factor of 3. To maximize detection throughput, we also run on multiple GPUs, realizing 95.6 FPS on 5 Fermi GPUs.     

Optimal production scheduling for hybrid manufacturing–remanufacturing systems with setups

Hybrid systems that use both raw materials (manufacturing mode) and returned products (remanufacturing mode) in their production process are considered. The system consists of one facility and necessitates setup for switching from one production mode to another. Since the flow rate of returned products is limited (fixed percentage of the demand rate is considered), switching from one mode to another is unavoidable, and so production and setup scheduling becomes critical for meeting customer demand and manufacturing cost optimization. Analytical solutions for production and setup strategies are obtained, feasibility conditions are derived, and the sensitivity of obtained results over system parameters is investigated. It is demonstrated that there exist two types of systems: mainly manufacturing systems with a relatively low rate of return, and mainly remanufacturing systems with a relatively low use of raw materials. Quantitative criteria distinguishing these two types of systems are developed, and it is shown that systems of different types obey different feasibility conditions and exhibit different optimal behavior.     

A novel hybrid PSO–GA meta-heuristic for scheduling of DAG with communication on multiprocessor systems

This work presents a novel hybrid meta-heuristic that combines particle swarm optimization and genetic algorithm (PSO–GA) for the job/tasks in the form of directed acyclic graph (DAG) exhibiting inter-task communication. The proposed meta-heuristic starts with PSO and enters into GA when local best result from PSO is obtained. Thus, the proposed PSO–GA meta-heuristic is different than other such hybrid meta-heuristics as it aims at improving the solution obtained by PSO using GA. In the proposed meta-heuristic, PSO is used to provide diversification while GA is used to provide intensification. The PSO–GA is tested for task scheduling on two standard well-known linear algebra problems: LU decomposition and Gauss–Jordan elimination. It is also compared with other states-of-the-art heuristics for known solutions. Furthermore, its effectiveness is evaluated on few large sizes of random task graphs. Comparative study of the proposed PSO-GA with other heuristics depicts that the PSO–GA performs quite effectively for multiprocessor DAG scheduling problem.     

Tardiness bounds under global EDF scheduling on a multiprocessor

We consider the scheduling of a sporadic real-time task system on an identical multiprocessor. Though Pfair algorithms are theoretically optimal for such task systems, in practice, their runtime overheads can significantly reduce the amount of useful work that is accomplished. On the other hand, if all deadlines need to be met, then every known non-Pfair algorithm requires restrictions on total system utilization that can approach approximately 50% of the available processing capacity. This may be overkill for soft real-time systems, which can tolerate occasional or bounded deadline misses (i.e. bounded tardiness). In this paper we derive tardiness bounds under preemptive and non-preemptive global when the total system utilization is not restricted, except that it not exceed the available processing capacity. Hence, processor utilization can be improved for soft real-time systems on multiprocessors. Our tardiness bounds depend on the total system utilization and per-task utilizations and execution costs—the lower these values, the lower the tardiness bounds. As a final remark, we note that global may be superior to partitioned for multiprocessor-based soft real-time systems in that the latter does not offer any scope to improve system utilization even if bounded tardiness can be tolerated.

Mobile agent-enabled framework for structuring and building distributed systems on the internet

Mobile agent has shown its promise as a powerful means to complement and enhance existing technology in various application areas. In particular, existing work has demonstrated that MA can simplify the development and improve the performance of certain classes of distributed applications, especially for those running on a wide-area, heterogeneous, and dynamic networking environment like the Internet. In our previous work, we extended the application of MA to the design of distributed control functions, which require the maintenance of logical relationship among and/or coordination of proc- essing entities in a distributed system. A novel framework is presented for structuring and building distributed systems, which use cooperating mobile agents as an aid to carry out coordination and cooperation tasks in distributed systems. The framework has been used for designing various distributed control functions such as load balancing and mutual ex- clusion in our previous work. In this paper, we use the framework to propose a novel ap- proach to detecting deadlocks in distributed system by using mobile agents, which dem- onstrates the advantage of being adaptive and flexible of mobile agents. We first describe the MAEDD (Mobile Agent Enabled Deadlock Detection) scheme, in which mobile agents are dispatched to collect and analyze deadlock information distributed across the network sites and, based on the analysis, to detect and resolve deadlocks. Then the design of an adaptive hybrid algorithm derived from the framework is presented. The algorithm can dynamically adapt itself to the changes in system state by using different deadlock detec- tion strategies. The performance of the proposed algorithm has been evaluated using simulations. The results show that the algorithm can outperform existing algorithms that use a fixed deadlock detection strategy.     

A scheduling algorithm for maximum throughput based on the link condition in heterogeneous network

Nowadays, a lot of wireless interfaces can be used by mobile users to access the Internet, such as WLAN, WiMAX, WlFI and even 3G. If a mobile terminal is equipped with multiple interfaces, it can use them simultaneously to improve the performance at the hot point where different RANs （Radio Access Networks） overlap. This paper proposes a scheduling algorithm based on the link condition that ensures the whole network has the maximum throughput. Simulation is also done to show the improvement of throughput with this scheduling algorithm.     

A distributed model predictive control scheme for leader–follower multi-agent systems

In this paper, we present a novel receding horizon control scheme for solving the formation problem of leader–follower configurations. The algorithm is based on set-theoretic ideas and is tuned for agents described by linear time-invariant (LTI) systems subject to input and state constraints. The novelty of the proposed framework relies on the capability to jointly use sequences of one-step controllable sets and polyhedral piecewise state-space partitions in order to online apply the ‘better’ control action in a distributed receding horizon fashion. Moreover, we prove that the design of both robust positively invariant sets and one-step-ahead controllable regions is achieved in a distributed sense. Simulations and numerical comparisons with respect to centralised and local-based strategies are finally performed on a group of mobile robots to demonstrate the effectiveness of the proposed control strategy.     

An efficient hybrid learning algorithm for neural network–based speech recognition systems on FPGA chip

This paper implemented an artificial neural network (ANN) on a field programmable gate array (FPGA) chip for Mandarin speech measurement and recognition of nonspecific speaker. A three-layer hybrid learning algorithm (HLA), which combines genetic algorithm (GA) and steepest descent method, was proposed to fulfill a faster global search of optimal weights in ANN. Some other popular evolutionary algorithms, such as differential evolution, particle swarm optimization and improve GA, were compared to the proposed HLA. It can be seen that the proposed HLA algorithm outperforms the other algorithms. Finally, the designed system was implemented on an FPGA chip with an SOC architecture to measure and recognize the speech signals.     

An effective hybrid teaching–learning-based optimization algorithm for permutation flow shop scheduling problem

Permutation flow shop scheduling (PFSP) is among the most studied scheduling settings. In this paper, a hybrid Teaching–Learning-Based Optimization algorithm (HTLBO), which combines a novel teaching–learning-based optimization algorithm for solution evolution and a variable neighborhood search (VNS) for fast solution improvement, is proposed for PFSP to determine the job sequence with minimization of makespan criterion and minimization of maximum lateness criterion, respectively. To convert the individual to the job permutation, a largest order value (LOV) rule is utilized. Furthermore, a simulated annealing (SA) is adopted as the local search method of VNS after the shaking procedure. Experimental comparisons over public PFSP test instances with other competitive algorithms show the effectiveness of the proposed algorithm. For the DMU problems, 19 new upper bounds are obtained for the instances with makespan criterion and 88 new upper bounds are obtained for the instances with maximum lateness criterion.