Module 06 / 07

Timing constraints are independent: task, reaction, sequence.

The framework must pass three independent timing checks: local task deadlines, the end-to-end response budget, and the minimum dwell time between reconfigurations.

Task, reaction, sequenceExpand section

Three independent timing conditions

Task, reaction, sequence

The framework must satisfy three different timing requirements: task schedulability, the end-to-end fault response budget, and the minimum interval between repeated reconfigurations.

R_i ≤ D_i
θ + R_dec + t_hot ≤ t_max(mode)
Δt_reconf ≥ τd

Dissertation source: §§4.2.5, 4.10.1–4.10.7

What was obtained for the reference implementationExpand section

Timing verification results

What was obtained for the reference implementation

For the framework task set, the total utilization is U = 0.396 with the bound at 0.735, and the worst response-time ratio R/D is only 0.176. The task set is therefore schedulable with a healthy reserve.

The end-to-end response consists of detection, decision and hot switching. The baseline result is t_reaction = 82.30 ms, while the best achievable value after optimization is 54.80 ms.

Dissertation source: §§4.2.5, 4.10.1–4.10.7

The bottleneck is not DecisionEngine but switchingExpand section

Design requirement

The bottleneck is not DecisionEngine but switching

The key conclusion is that the limiting factor is not configuration selection but the transition mechanism t_hot. To satisfy the DEGRADED-mode requirement, hot switching must be reduced to t_hot ≤ 30.2 ms.

Dissertation source: §§4.2.5, 4.10.1–4.10.7

Back to overview →Next →The planner must choose a trajectory, not only the next node.