Neural Feedback Scheduling of Real-Time Control Tasks

Many embedded real-time control systems suffer from resource constraints and dynamic workload variations. Although optimal feedback scheduling schemes are in principle capable of maximizing the overall control performance of multitasking control systems, most of them induce excessively large computational overheads associated with the mathematical optimization routines involved and hence are not directly applicable to practical systems. To optimize the overall control performance while minimizing the overhead of feedback scheduling, this paper proposes an efficient feedback scheduling scheme based on feedforward neural networks. Using the optimal solutions obtained offline by mathematical optimization methods, a back-propagation (BP) neural network is designed to adapt online the sampling periods of concurrent control tasks with respect to changes in computing resource availability. Numerical simulation results show that the proposed scheme can reduce the computational overhead significantly while delivering almost the same overall control performance as compared to optimal feedback scheduling.

💡 Research Summary

The paper addresses a fundamental challenge in embedded real‑time control systems: how to allocate limited CPU resources among several concurrent control tasks while coping with dynamic workload variations. Traditional optimal feedback scheduling formulates this as a constrained nonlinear optimization problem that minimizes a weighted sum of control costs J = ∑ w_i J_i(h_i) subject to the schedulability constraint ∑ c_i/h_i ≤ U_R, where h_i denotes the sampling period of task i, c_i its execution time, and U_R the desired CPU utilization. Solving this problem online with methods such as Sequential Quadratic Programming (SQP) yields high‑quality schedules but incurs prohibitive computational overhead because each iteration requires gradient, Hessian, and quadratic‑program sub‑solves. Consequently, such optimal schedulers are rarely usable in practice.

The authors propose a Neural Feedback Scheduler (NFS) that replaces the online optimizer with a feed‑forward back‑propagation (BP) neural network. The key idea is to pre‑compute a large set of optimal solutions offline using SQP for a wide range of possible inputs (the vector of execution times c =