Many-core acceleration for model predictive control systems

Abstract

This paper proposes a novel many-core execution strategy for real-time model predictive controls. The key idea is to exploit predicted input values, which are produced by the model predictive control itself, to speculatively solve an optimal control problem. It is well known that control applications are not suitable for multi- or many-core processors, because feedback-loop systems inherently stand on sequential operations. Since the proposed scheme does not rely on conventional thread-/data-level parallelism, it can be easily applied to such control systems. An analytical evaluation using a real application demonstrates the potential of performance improvement achieved by the proposed speculative executions.