Relaxing Signal Delay Constraints in Distributed Embedded Controllers

Abstract

Abstract-Embedded systems often involve transmitting feedback signals between multiple control tasks that are implemented on different ECUs communicating via a shared bus. For ensuring stability and control performance, such designs require all control signals to be delivered within a specified deadline, which is ensured through appropriate timing or schedulability analysis. In this paper we study controller design that allows control feedback signals to occasionally miss their deadlines. In particular, we provide analytical bounds on deadline misses such that the control loop retains its stability and meets its control performance requirements. We argue that such relaxation allows us to (i) use lower quality communication resources (e.g., event-triggered instead of time-triggered communication) and (ii) provide more flexibility -e.g., use simulation -in communication timing analysis since analytical worst-case delay bounds for real-life communication protocols are often pessimistic. We illustrate this approach using the FlexRay communication protocol for distributed automotive control systems.

Composable Platform-Aware Embedded Control Systems on a Multi-core Architecture

“…We illustrate the applicability of our proposed platformaware design flow considering an automotive cruise control system [14]. The purpose of the system is to maintain a constant vehicle speed despite external disturbances.…”

Effectively utilizing elastic resources in networked control systems

“…However, most of the design approaches, ensuring safety and worst-case performance, impose constraints on network determinism such as maximum rate of packet drops [2], [3], [4], worst-case delay and jitter [5]. Otherwise, if the communication timings are stochastic, then only probabilistic or approximate analysis is possible [13], [14], which is not acceptable for safety-critical systems.…”

“…For example, the maximum rate of packet drops is assumed in [2], [3], [4]. The worst-case delay and jitter are assumed in [5].…”

Automotive Cyber–Physical Systems: A Tutorial Introduction

“…One of the issues here is to quantify the tolerable message loss/delay in the case of distributed controller implementations, while still maintaining control quality [27]. Once such tolerable loss patterns are identified, the next issue is to design scheduling policies that respect such bounds or patterns.…”