Greedy Alignment

Recently, scheduling problems proceed in domains as diverse as manufacturing, computer processing, transportation, health care, space exploration, and education. A scheduling technique role is to organize the allocation of resource(s) over time to perform allocation of tasks. This paper presents the design and analysis of a generalized Job-Shop (JS) problem with priority assignment of jobs. Two Feed-Forward Neural Networks (FFNN) have been developed to solve the scheduling problem. The starting time for various operations will be determined using FFNN. Then, these operations will be scheduled by greedy algorithm. The greedy alignment procedure aligns operations of various jobs on respective machines with feasible/optimal Finishing Time (FT). Each operation within a job is predefined priority by another FFNN, which is trained to recognize jobs from a job queue to estimate each job’s priority. Simulation results prove the validity of the proposed approach.

A real-time operating system (RTOs) is often used in embedded system, to structure the application code and to ensure that the deadlines are met by reacting on events by executing the functions within precise time. Most embedded systems are bound to real-time constraints with determinism and latency as a critical metrics. RTOs are generally implemented in software, increases computational overheads, jitter and memory footprint. Modern FPGA technology, enables the implementation of a full featured and flexible hardware based RTOs, which helps in reducing to greater extent these overheads even if not remove completely. Scheduling algorithms play an important role in the design of real-time systems. An Adaptive Fuzzy Inference System (FIS) based scheduler framework proposed in this article is based on the study and conclusion drawn from the research over the years in HW SW co-design domain. The proposed novel two phase FIS based adaptive hardware task scheduler minimizes the processor time for scheduling activity which uses fuzzy logic to model the uncertainty at first stage along with adaptive framework that uses feedback which allows processors share of task running on multiprocessor to be controlled dynamically at runtime. This Fuzzy logic based adaptive hardware scheduler breakthroughs the limit of the number of total task and thus improves efficiency of the entire real-time system. The increased computation overheads resulted from proposed two phase FIS scheduler can be compensated by utilising the basic characteristics of parallelism of the hardware as scheduler being migrated to FPGA.

A real-time operating system (RTOs) is often used in embedded system, to structure the application code and to ensure that the deadlines are met by reacting on events by executing the functions within precise time. Most embedded systems are bound to real-time constraints with determinism and latency as a critical metrics. RTOs are generally implemented in software, increases computational overheads, jitter and memory footprint. Modern FPGA technology, enables the implementation of a full featured and flexible hardware based RTOs, which helps in reducing to greater extent these overheads even if not remove completely. Scheduling algorithms play an important role in the design of real-time systems. An Adaptive Fuzzy Inference System (FIS) based scheduler framework proposed in this article is based on the study and conclusion drawn from the research over the years in HW SW co-design domain. The proposed novel two phase FIS based adaptive hardware task scheduler minimizes the processor time for scheduling activity which uses fuzzy logic to model the uncertainty at first stage along with adaptive framework that uses feedback which allows processors share of task running on multiprocessor to be controlled dynamically at runtime. This Fuzzy logic based adaptive hardware scheduler breakthroughs the limit of the number of total task and thus improves efficiency of the entire real-time system. The increased computation overheads resulted from proposed two phase FIS scheduler can be compensated by utilising the basic characteristics of parallelism of the hardware as scheduler being migrated to FPGA.

Job-Shop Scheduling (JSS) processes have highly complex structure in terms of many criteria. Because there is no
limitation in the number of the process and there are many alternative scheduling. In JSS, each order that is processed
on different machines has its own process and process order. It is very important to put these processes into a sequence
according to a certain order. In addition, some constraints must be considered in order to obtain the appropriate tables.
In this paper, a 3-layers Feed Forward Backpropagation Neural Network (FFBNN) has been used for two different
purposes, the first one task is to obtain the priority and the second one role is to determine the starting order of each
operation within a job. Precedence order of operations indicates the dependency of subtasks within a job, Furthermore,
the combined greedy procedure along with the back propagation algorithm will align operations of each job until best
solution is obtained. In particular, greedy type algorithm will not always find the optimal solution. However, adding
a predefined alignment dataset along with the greedy procedure result in optimal solutions.

Recently, scheduling problems proceed in domains as diverse as manufacturing, computer processing, transportation, health care, space exploration, and education. A scheduling technique role is to organize the allocation of resource(s) over time to perform allocation of tasks. This paper presents the design and analysis of a generalized Job-Shop (JS) problem with priority assignment of jobs. Two Feed-Forward Neural Networks (FFNN) have been developed to solve the scheduling problem. The starting time for various operations will be determined using FFNN. Then, these operations will be scheduled by greedy algorithm. The greedy alignment procedure aligns operations of various jobs on respective machines with feasible/optimal Finishing Time (FT). Each operation within a job is predefined priority by another FFNN, which is trained to recognize jobs from a job queue to estimate each job’s priority. Simulation results prove the validity of the proposed approach.