Ensuring End-to-End Latencies of Multi-Rate Cause-Effect Chains in Multi-Mode Systems Through Migration

Safety-critical control applications running on multi-core platforms often need to operate under multiple execution modes during runtime. A transition from one mode to another may cause the task set to change, e.g., new tasks join the system, which might increase the end-to-end latencies of certain cause-effect chains and cause them to violate their latency requirements. In this paper, we propose a method to enhance the feasibility of multi-rate cause-effect chains that apply the Logical Execution Time (LET) model to meet end-to-end latency requirements in multi-mode systems. Specifically, our method selectively migrates task instances to enable the reconfiguration of LET-based communication intervals. Results using an automotive benchmark and synthetic task sets show that even in scenarios with high utilization of the cores, our method can improve the end-to-end latencies guarantees of cause-effect chains.