Heuristic Scheduling Research Articles

Instance-Rotation-Based Surrogate in Genetic Programming With Brood Recombination for Dynamic Job-Shop Scheduling

Genetic programming has achieved great success for learning scheduling heuristics in dynamic job shop scheduling. In theory, generating a large number of offspring for genetic programming, known as brood recombination, can improve its heuristic generation ability. However, it is time-consuming to evaluate extra individuals. Phenotypic characterisation based surrogates with K-nearest neighbours have been successfully used for genetic programming to preselect only promising individuals for real fitness evaluations in dynamic job shop scheduling. However, sample individuals used by surrogate are from only the current generation, since the fitness of individuals across generations are not comparable due to the rotation of training instances. The surrogate cannot accurately estimate the fitness of an offspring that is far away from all the limited sample individuals at the current generation. This paper proposes an effective instance rotation based surrogate to address the above issue. Specifically, the surrogate uses the samples extracted from individuals across multiple generations with different instances. More importantly, we propose a fitness mapping strategy to make the fitness evaluated by different instances comparable. The results show that the GP with brood recombination and the proposed surrogate can significantly improve the quality of scheduling heuristics. The results also reveal that the proposed algorithm has successfully reduced the number of omitted promising offspring due to the higher accuracy of the surrogate. The samples in the new surrogate spread better in the phenotypic space, and the nearest neighbour tends to be closer to the predicted offspring. This makes the estimated fitness more accurate.

DTRL: Decision Tree-based Multi-Objective Reinforcement Learning for Runtime Task Scheduling in Domain-Specific System-on-Chips

Domain-specific systems-on-chip (DSSoCs) combine general-purpose processors and specialized hardware accelerators to improve performance and energy efficiency for a specific domain. The optimal allocation of tasks to processing elements (PEs) with minimal runtime overheads is crucial to achieving this potential. However, this problem remains challenging as prior approaches suffer from non-optimal scheduling decisions or significant runtime overheads. Moreover, existing techniques focus on a single optimization objective, such as maximizing performance. This work proposes DTRL, a decision-tree-based multi-objective reinforcement learning technique for runtime task scheduling in DSSoCs. DTRL trains a single global differentiable decision tree (DDT) policy that covers the entire objective space quantified by a preference vector. Our extensive experimental evaluations using our novel reinforcement learning environment demonstrate that DTRL captures the trade-off between execution time and power consumption, thereby generating a Pareto set of solutions using a single policy. Furthermore, comparison with state-of-the-art heuristic–, optimization–, and machine learning-based schedulers shows that DTRL achieves up to 9× higher performance and up to 3.08× reduction in energy consumption. The trained DDT policy achieves 120 ns inference latency on Xilinx Zynq ZCU102 FPGA at 1.2 GHz, resulting in negligible runtime overheads. Evaluation on the same hardware shows that DTRL achieves up to 16% higher performance than a state-of-the-art heuristic scheduler.

Energy-Efficient Service-Aware Multi-Connectivity Scheduler for Uplink Multi-Layer Non-Terrestrial Networks

This paper introduces the concept of energy efficiency (EE) in the uplink with the capability of multi-connectivity (MC) in a multi-orbit non-terrestrial network (NTN), where user terminals (UTs) can be simultaneously served by more than one satellite to achieve higher peak throughput at reduced energy consumption. This concept also considers the service classification of the users, so that network dimensioning is performed in order to satisfy the quality of service (QoS) requirement of users. MC can increase throughput, but this entails increased power consumption at user terminal for uplink transmissions. To this end, an energy-efficient service-aware multi-connectivity (EE-SAMC) scheduling algorithm is developed in this paper to improve the EE of uplink communications. EE-SAMC uses available radio resources and propagation information to intelligently define a dynamic resource allocation pattern, that optimally routes traffic so as to reduce the energy consumption at the UT while ensuring QoS is maximized. EE-SAMC is designed based on the formulation of a non-convex combinatorial problem, it is solved in two ways involving firstly an optimization solution and secondly a heuristic approach. The effectiveness of EE-SAMC is compared with random allocation, round robin and heuristic schedulers in terms of EE, throughput and delay; EE-SAMC outperforms all schedulers.

Learning to schedule heuristics for the simultaneous stochastic optimization of mining complexes

The simultaneous stochastic optimization of mining complexes is a large-scale stochastic combinatorial optimization problem that simultaneously manages the extraction of materials from multiple mines and their processing through interconnected facilities to generate a set of final products, while taking into account geological (material supply) uncertainty to manage the associated risk. Existing methods do not offer solutions to such a complex problem in a reasonable time. This work proposes a data-driven framework for heuristic scheduling in a fully self-managed hyper-heuristic to solve the simultaneous stochastic optimization of mining complexes. The proposed learn-to-perturb hyper-heuristic is a multi-neighborhood simulated annealing algorithm that selects the heuristic (perturbation) to be applied in a self-adaptive manner using reinforcement learning to efficiently explore the local search that is best suited to a particular search point. By learning from data that describes the performance of the heuristics, a problem-specific ordering of heuristics that collectively find better solutions faster is obtained. To the best of our knowledge, this research proposes the first data-driven heuristic search tree for mine planning. Results from several instances of two types of large-scale industrial mining complexes show a reduction of up to 80% in execution time and an order of magnitude reduction in primal suboptimality compared to state-of-the-art methods.

A priority rule for scheduling shared due dates in the resource-constrained project scheduling problem

In resource-constrained project scheduling, project milestones, rental equipment, contracted labor, and product development checkpoints all conceptually impose shared due dates, which expand the concept of conventional due dates by permitting their assignment to sets of activities and by allowing multiple shared due date assignments, or none at all, to a single activity. Existing heuristic algorithms in the literature that solve the resource-constrained project scheduling problem are agnostic to shared due dates and cannot utilize information about their assignments in making scheduling decisions. To this end, we propose a novel priority rule compatible with both serial and parallel schedule generation schemes that integrates shared due date information into the heuristic scheduling decisions. We evaluate the effectiveness of our priority rule by demonstrating its dominance over existing priority rules in the literature and by comparing its performance with an exact algorithm. In comparing with the latter, we demonstrate both the computational efficiency of the heuristic as well as the quality of the solutions it produces.

Multitask Multiobjective Genetic Programming for Automated Scheduling Heuristic Learning in Dynamic Flexible Job-Shop Scheduling.

Evolutionary multitask multiobjective learning has been widely used for handling more than one multiobjective task simultaneously. However, it is rarely used in dynamic combinatorial optimization problems, which have valuable practical applications such as dynamic flexible job-shop scheduling (DFJSS) in manufacturing. Genetic programming (GP), as a popular hyperheuristic approach, has been used to learn scheduling heuristics for generating schedules for multitask single-objective DFJSS only. Searching in the heuristic space with GP is more difficult than in the solution space, since a small change on heuristics can lead to ineffective or even infeasible solutions. Multiobjective DFJSS is more challenging than single DFJSS, since a scheduling heuristic needs to cope with multiple objectives. To tackle this challenge, we first propose a multipopulation-based multitask multiobjective GP algorithm to preserve the quality of the learned scheduling heuristics for each task. Furthermore, we develop a multitask multiobjective GP algorithm with a task-oriented knowledge-sharing strategy to further improve the effectiveness of learning scheduling heuristics for DFJSS. The results show that the designed multipopulation-based GP algorithms, especially the one with the task-oriented knowledge-sharing strategy, can achieve good performance for all the examined tasks by maintaining the quality and diversity of individuals for corresponding tasks well. The learned Pareto fronts also show that the GP algorithm with task-oriented knowledge-sharing strategy can learn competitive scheduling heuristics for DFJSS on both of the objectives.

Joint Scheduling of Participants, Local Iterations, and Radio Resources for Fair Federated Learning over Mobile Edge Networks

During the training process of Federated learning (FL), proper devices are selected to participate in the training process to avoid model unfairness. In a mobile edge network, participant selection must be considered together with three factors: non-iid datasets possessed by devices, tunable local iterations on devices, and radio resource allocation to counter the impact of time-varying channel conditions on parameter transmissions. Since datasets of devices are given, to ensure model fairness and achieve fast convergence in the FL training process, participants, local iterations, and radio resources must be scheduled jointly in each iteration of FL training. In this paper, the joint scheduling problem is analyzed and formulated. Since it is NP-hard, a heuristic scheduling method called PALORA is designed to conduct joint scheduling of participants, local iterations, and radio resources. PALORA consists of three sequentially interactive function blocks: 1) a pointer network embedded deep reinforcement learning method to select participants, 2) an estimation algorithm to determine the numbers of local iterations, and 3) a breadth-first search method to allocate radio resources to the selected participants. PALORA is evaluated via extensive simulations based on real-world datasets. Results show that it significantly outperforms benchmark approaches.

FedRich: Towards efficient federated learning for heterogeneous clients using heuristic scheduling

Federated Learning (FL) always involves a large number of heterogeneous clients (both in statistic and resource heterogeneity) when collaboratively training a model, leading to a compromise in the model performance. Recent research has focused on customizing FL frameworks to address the issues. However, compared to models trained on independent and identically distributed (IID) data, these methods still face performance degradation in non-IID scenarios. Moreover, resource consumption is also somewhat expensive. In this work, we present an efficient FL framework named FedRich to tackle the statistic and resource heterogeneity. The key idea of FedRich is adaptive segmentation of the model and heuristic scheduling of the active clients. Adaptive segmentation enables resource-dependent customization of the model, which is conducive to clients with varying resource budgets to conduct local training. The heuristic scheduling strategy appropriately selects clients to participate in federated training, mitigating statistical heterogeneity. Moreover, FedRich incorporates a hierarchical aggregation mechanism to stably aggregate heterogeneous models of different sophistication. Extensive experimental results on three benchmark datasets demonstrate that FedRich outperforms state-of-the-art heterogeneous FL approaches.

Time-aware deterministic bandwidth allocation scheme in TDM-PON for time-sensitive industrial flows

A programmable logic controller deployed in the cloud or edge cloud (cloud PLC) contributes to flexible and scalable industrial manufacturing processes. For the industrial Internet, a time-division multiplexing passive optical network is a promising access technology for connecting cloud PLC and industrial devices. Time-sensitive (TS) industrial applications such as motion control and mobile robotics impose strict delay and jitter requirements, pursuing deterministic transmission capability, that is, guaranteeing that the upper bounds on delay and jitter of the TS industrial flow meet its requirements. However, traditional bandwidth allocation schemes aim to leverage statistical multiplexing to improve bandwidth resource utilization efficiency, so it guarantees only the “average delay/jitter” of the service and cannot meet the deterministic requirements of the TS industrial flow. This paper presents a time-aware deterministic bandwidth allocation (TA-DetBA) scheme to satisfy the deterministic transmission of TS industrial flows. To this purpose, first, we constrain the position of the transmission window (i.e., bandwidth) allocated by optical line termination to the optical network unit according to the arrival characteristics (i.e., flow arrival time and flow cycle) and requirements (i.e., delay and jitter) of the flow. Subsequently, we propose an integer linear programming scheduling model and a greedy-based time slot equalization heuristic scheduling algorithm to improve schedulability and reduce the average delay of best-effort flows. Simulation results show that our proposed TA-DetBA scheme guarantees the deterministic delay and jitter of TS industrial flow. Compared with the fixed bandwidth allocation (FBA) scheme, the average resource utilization efficiency of the TA-DetBA scheme is improved by up to 20.6% higher than FBA.

A Two-stage Multi-population Genetic Algorithm with Heuristics for Workflow Scheduling in Heterogeneous Distributed Computing Environments

Workflow scheduling in Heterogeneous Distributed Computing Environments (HDCEs) is a NP-hard problem. Although a number of scheduling approaches have been proposed for workflow scheduling in HDCEs, there is still a room and need for improvement. To fill the gaps, this study formulates workflow scheduling problem in HDCEs as a complete, solvable and extensible integer programming mathematical model with precedence and resource constraints that provides a theoretical foundation for developing workflow scheduling strategy. Then, this study develops a novel two-stage multi-population genetic algorithm with heuristics for workflow scheduling. In particular, two-stage multi-population coevolution strategy is employed with designed novel methods for population initialization, genetic operation, individual decoding and improvement. To estimate the validity, extensive experiments are designed and conducted on various scenarios based on real and random workflow applications. The results have shown the practical viability of the proposed algorithm outperforming conventional approaches.

Reasoning-Based Scheduling Method for Agile Earth Observation Satellite with Multi-Subsystem Coupling

With the rapid development of agile Earth observation satellites (AEOSs), these satellites are able to conduct more high-quality observation missions. Nevertheless, while completing these missions takes up more data transmission and electrical energy resources, it also increases the coupling within each satellite subsystem. To address this problem, we propose a reasoning-based scheduling method for an AEOS under multiple subsystem constraints. First, we defined the AEOS mission scheduling model with multi-subsystem constraints. Second, we put forward a state variable prediction method that reflects the different coupling states of a satellite after analyzing the coupling relationships between various subsystems and identifying the primary limiting coupling states for each subsystem. Third, we established the reasoning rules corresponding to the planning strategies of different coupling states of the satellite by adding two planning strategies based on the planning strategies of existing planning methods. By comparing the proposed method to three heuristic scheduling methods and a meta-heuristic scheduling method, the results show that our method has better performance in terms of scheduling results and efficiency.

Job scheduling for maximum revenue on uniform, parallel machines with major and minor setups and job splitting

We consider a revenue maximization scheduling problem where jobs can be split on parallel, uniform machines. By assuming that only major and minor setups exist between jobs and that these can be grouped into families, we reformulate the scheduling problem as a continuous knapsack problem with setup time. We then create the Revenue Rate Heuristic to solve the maximum revenue scheduling problem for both the partial and binary revenue cases. This makes use of a revenue rate – the ratio between total revenue that can be obtained from a job and the time required to achieve the revenue. The Revenue Rate Heuristic schedules the job with the highest revenue rate first and updates all the revenue rates accordingly based on existing scheduled families and remaining capacity. In cases where there are no setup times or if there is only one machine and an ordering of job revenue and production time, we are able to show that the Revenue Rate Heuristic is optimal. Numerical studies show that the Revenue Rate Heuristic performs well with an average optimality gap of 3.47% across a large set of parameters and when the setup times across families are different from each other. In a case study based on firm-level data, we also show that the average optimality gap remains small.

Machine Learning Feature Based Job Scheduling for Distributed Machine Learning Clusters

With the rapid proliferation of Machine Learning (ML) and Deep learning (DL) applications running on modern platforms, it is crucial to satisfy application performance requirements such as meeting deadline and ensuring accuracy. To this end, researchers have proposed several job schedulers for ML clusters. However, none of the previously proposed schedulers consider ML model parallelism, though it has been proposed as an approach to increase the efficiency of running large-scale ML and DL jobs. Thus, in this paper, we propose an ML job Feature based job Scheduling system (MLFS) for ML clusters running both data parallelism and model parallelism ML jobs. MLFS first uses a heuristic scheduling method that considers an ML job’s spatial and temporal features to determine task priority for job queue ordering in order to improve job completion time (JCT) and accuracy performance. It uses the data from the heuristic scheduling method for training a deep reinforcement learning (RL) model. After the RL model is well trained, it then switches to the RL method to automatically make decisions on job scheduling. In addition, MLFS has a system load control method that selects tasks from overloaded servers to move to underloaded servers based on task priority, and also intelligently removes the tasks that generate little or no improvement on the desired accuracy performance when the system is overloaded to improve JCT and accuracy by job deadline. Furthermore, we propose Optimal ML iteration stopping method that determines the proper time to stop training ML model when this model reaches the minimum loss value. Our real experiments and large-scale simulation based on real trace show that MLFS reduces JCT by up to 53% and makespan by up to 52%, and improves accuracy by up to 64% when compared with existing ML job schedulers. We also open sourced our code.

A New Algorithm for Real-Time Scheduling and Resource Mapping for Robot Operating Systems (ROS)

A new version of a robot operating system (ROS-2) has been developed to address the real-time and fault constraints of distributed robotics applications. However, current implementations lack strong real-time scheduling and the optimization of response time for various tasks and applications. This may lead to inconsistent system behavior and may affect system performance. This article presents a new and efficient heuristic scheduling algorithm for the ROS-2 framework to improve resource mapping and system scheduling for robotics applications. The proposed scheduling design includes grouping callbacks into executors, assigning executor priority, and sequencing callbacks inside executors. The timing constraints and functional properties are expressed in the formal design model, and real-time scheduling is performed with respect to the timing constraints. The benefits of the proposed solution are demonstrated by a set of experimental results. Using this algorithm, it is shown that the number of executors needed to schedule a set of callbacks is minimized and the system schedulability is maximized for loaded and overloaded cases.

Heuristics for parallel machine scheduling with GoS eligibility constraints

Heuristics for parallel machine scheduling with GoS eligibility constraints

Rules for selecting orders from the precedence in production systems

Analytical rules for selecting orders from the precedence by using heuristic scheduling algorithms in MES and APS class systems are considered. The proposed sampling rules are presented for single-stage and multistage planning systems, subjected to the intensity of orders. Keywords heuristic scheduling algorithm, order intensity coefficient, one-stage systems, multi-stage systems, slack time, directive due dates

An actor-critic framework based on deep reinforcement learning for addressing flexible job shop scheduling problems.

With the rise of Industry 4.0, manufacturing is shifting towards customization and flexibility, presenting new challenges to meet rapidly evolving market and customer needs. To address these challenges, this paper suggests a novel approach to address flexible job shop scheduling problems (FJSPs) through reinforcement learning (RL). This method utilizes an actor-critic architecture that merges value-based and policy-based approaches. The actor generates deterministic policies, while the critic evaluates policies and guides the actor to achieve the most optimal policy. To construct the Markov decision process, a comprehensive feature set was utilized to accurately represent the system's state, and eight sets of actions were designed, inspired by traditional scheduling rules. The formulation of rewards indirectly measures the effectiveness of actions, promoting strategies that minimize job completion times and enhance adherence to scheduling constraints. The experimental evaluation conducted a thorough assessment of the proposed reinforcement learning framework through simulations on standard FJSP benchmarks, comparing the proposed method against several well-known heuristic scheduling rules, related RL algorithms and intelligent algorithms. The results indicate that the proposed method consistently outperforms traditional approaches and exhibits exceptional adaptability and efficiency, particularly in large-scale datasets.

ИНТЕГРИРОВАННАЯ МОДЕЛЬ РЕШЕНИЯ ЗАДАЧИ ДИСПЕТЧЕРИЗАЦИИ ЗАЯВОК

The paper considers the problem of scheduling. The paradigm of organization of distributedcomputing based on Grid-computing is considered. The classification of task scheduling systems isgiven. Various approaches to solving the scheduling problem are described. A model of the task ofservicing applications based on the principles of the theory of queuing systems is presented. Thetask statement is formulated on the basis of Grid-scheduling. The concept of a resource rectangleis proposed. The environment for scheduling resource rectangles is defined. A model is proposedthat allows formalizing the user's request for service by the concept of a resource (non-Euclidean)rectangle. Instead of the principle of optimization based on the machine search for the best distributionof the array of resource rectangles, a heuristic principle was proposed, which made it possibleto reduce the amount of necessary calculations. The proposed heuristic scheduling algorithmmakes it possible to take into account the properties of the array and evaluate the quality of solutions.Models of the demand environment in the form of single cubic faces are constructed.The model of cubic faces is generalized to the experiment of cubic layers. The description of thedemand model used is given. A model of the resource supply environment in the form of a canonicalpyramid is constructed and the concept of a canonical demand-supply experiment for modelhomogeneous resource elements is introduced. A truncation of the supply-demand experiment hasbeen introduced. A hybrid model based on a combination of evolutionary search principles andfuzzy control methods is proposed. To solve scheduling problems, it is proposed to use evolutionaryalgorithms. A modified solution coding technique and new modifications of genetic operatorsfor solving scheduling problems have been developed. A block diagram of the algorithm for solvingthe problem under consideration is presented, taking into account the use of a fuzzy logic controller.Computer simulation has been performed and the results of computational experimentshave been presented. The features of the proposed method are revealed, its advantages and disadvantagesare formulated.

Multiple Objective Fairness Scheduling Optimization Algorithms Based on Multiple DAGs in Heterogeneous Edge Computing

The rapid development of edge computing has promoted the growing demand of applications. Quality of service (QoS) must be improved because this index involves issues such as unfairness and short Directed Acyclic Graph (DAG) schedule length. We propose a new algorithm named Multiple DAGs Depth Fairness Task Scheduling (MDDF) in this study to solve the multi-DAG task unfairness problem effectively, the algorithm on the basis of the measured depth of the task in the DAG, improves the fairness of multiple DAG, enhance the user QoS and obtain a fast response. The proposed algorithm is then compared with four common heuristic scheduling methods to verify its effectiveness. We propose another algorithm called Fast Multi-DAG Depth Fairness Task Scheduling (FMDDF) with high average slowdown to reduce the schedule length of short DAG and accelerate the response time. Experimental results show that FMDDF can improve scheduling and shorten the schedule length of short DAG by about 5% and MDDF can enhance fairness by about 2% while reducing the short DAG schedule length compared with other algorithms.

FATS-2TC: A Fault Tolerant real-time Scheduler for energy and temperature aware heterogeneous platforms with Two types of Cores

Fault tolerance in safety-critical real-time systems is indispensable. With the increasing density of transistors on MPSoCs, energy consumption and peak temperatures are exponentially increasing, thus making the method unreliable. This paper devises a heuristic scheduling strategy called FATS-2TC, which solves the issue of the simultaneous control of energy and peak temperature levels using the standby-sparing mechanism on systems with two types of cores (like big.LITTLE architecture) to make it resistant to transient faults. Our experimental results demonstrate that FATS-2TC can save considerable energy and lower core peak temperatures than a state-of-the-art energy-efficient fault-tolerant scheduler.