NaFlex Variant Scheduling

• NaFlex Variant is a deterministic multi-variant FlexRay static-segment scheduling formulation designed to synchronize shared signals across vehicle variants.
• The approach formalizes the scheduling problem using an exact mathematical model and employs a First-Fit heuristic with various signal ordering schemes.
• Empirical benchmarks demonstrate a 25–30% reduction in slot usage with advanced ordering methods like FFP and FFC, highlighting practical scalability.

The NaFlex Variant refers to the multi-variant, time-constrained FlexRay static-segment scheduling problem, where deterministic time-triggered communication schedules must be constructed for multiple vehicle variants sharing a platform. This formulation arises in automotive systems engineering where FlexRay buses, with zero jitter for static segment transmission, require simultaneous placement of shared signals across several vehicle variants to facilitate signal traceability, diagnostics, and component reuse. The approach formalizes the problem as an exact mathematical model, introduces an efficient First-Fit heuristic for schedule synthesis, and demonstrates scalability and optimality relative to standard benchmarks (Dvořák et al., 2020).

1. Formal Mathematical Model

The problem instance consists of the following elements:

• Signals $S$; Vehicle variants $V = \{1, \dots, m\}$; Static-segment cycle length $F$; Payload width $W$ (bits); Set of available slot indices $L = \{1, \dots, L_{\max}\}$; Hyperperiod of $C_{\max}$ communication cycles.
• For each signal $i \in S$: period $T_i$ (divides the hyperperiod), release date $O_i$, deadline $D_i$, payload size $C_i$, and transmitting node/ECU $N_i$.
• Variant-incidence matrix $V \in \{0,1\}^{|S| \times m}$, with $V_{i,j} = 1$ iff signal $i$ must appear in variant $j$.

The multi-schedule requirement stipulates that a signal used by multiple variants must occupy the same (cycle, slot, offset) in every schedule. Decision variables are $x_{i,c,\ell,o} \in \{0,1\}$, with $x_{i,c,\ell,o} = 1$ iff signal $i$'s first instance is placed in cycle $c$, slot $\ell$, bit-offset $o$. The objective is to minimize the total number of slots used in any communication cycle:

$$\min \left| \{ \ell \in L : \exists i, c, o \text{ such that } x_{i,c,\ell,o}=1 \} \right|$$

Major constraints include:

• Each signal must appear once within its time window: $$\sum_{c=r_i}^{d_i} \sum_{\ell \in L} \sum_{o=0}^{W-C_i} x_{i,c,\ell,o} = 1$$ for all $i \in S$, with $r_i = \lceil O_i / F \rceil$ and $d_i = \lfloor D_i / F \rfloor$.
• Frame payload capacity: non-overlapping bit intervals for co-occurring signals; otherwise only if those signals never co-occur in any variant, as encoded by the precomputed mutual-exclusion matrix $SMEM_{i,i'}$.
• ECU-slot compatibility enforced by node-exclusion matrix $NMEM_{n,n'}$.
• Time-window enforcement on placements.

2. Mutually Exclusive Signal and Node Conflicts

For any pair of signals $(i, i')$, mutual-exclusion is determined by $SMEM_{i,i'}$, which is set to $1$ iff there exists a variant where both signals appear, and $0$ otherwise. This ensures that two signals sharing at least one variant cannot overlap in the same frame. Similarly, node conflicts in slot assignments are excluded via the node-mutual-exclusion matrix $NMEM_{n,n'}$, which is calculated for all nodes and prohibits conflicting node assignments in the same slot for any variant.

3. First-Fit Heuristic and Signal Ordering Schemes

The NaFlex Variant algorithm uses a three-phase First-Fit heuristic:

1. Preprocessing: Compute the signal and node mutual-exclusion matrices and round timing constraints to cycle indices.
2. Signal Ordering: Create an ordered list $SL$ of signals, using strategies such as:
   • no ordering (FF)
   • increasing period $T_i$ (FFP)
   • increasing time window $d_i - r_i$ (FFW)
   • decreasing payload $C_i$ (FFL)
   • combined stable sort $( -C_i, d_i - r_i, T_i, N_i )$ (FFC)
3. Iterative Placement: For each signal $i$ in $SL$, search existing slots and cycles for a suitable position, checking ECU compatibility and available bit intervals using a bitmask. If no fit is available, allocate a new slot and assign the signal at the earliest cycle and offset.

The process continues until all signals are placed, ensuring simultaneity for shared signals across schedules.

4. Computational Complexity Analysis

Let $n = |S|$ be the number of signals, $s = L_{\text{alloc}}$ the allocated slot count, $H = C_{\max}/F$ the maximum cycles per signal. In the worst case, each signal tests $O(s \cdot H)$ frames, with offset search complexity $O(W + q)$ where $q$ is the number of signals already placed in the frame. The total runtime is $O(n \cdot s \cdot H \cdot W)$. Given that $s$, $H$, and $q$ remain small and bounded in practice, empirical observations show nearly linear scaling in $n$.

5. Experimental Validation and Benchmark Results

Benchmarking used extended SAE problem sets with 500–1 000 signals, 3–23 ECUs, payloads ranging from 32–128 bits, and 20 random variants per instance (signal inclusion probability $\leq 0.7$). The evaluation compared:

• TS (two-stage heuristic)
• FF, FFP, FFW, FFL, FFC (First-Fit variants)

Metrics were slot-count per cycle (averaged over 10 instances per set) and CPU runtime. Key outcomes:

• FFP and FFC algorithms consistently achieved the shortest schedules, producing a 25–30% reduction in slot count versus TS and unsorted FF.
• FFL (biggest-payload first) performed worst, attributed to strong correlation between payload size and long periods.
• FFP and FFW executed in $<$0.1s for up to 1 000 signals and in $<$1.2s for 3 000 signals; TS required up to 16 s in the 3 000-signal case.
• FFC ordering yielded the lowest slot count (within 1% of FFP) at moderate extra computational cost.

6. Practical Implications and Limitations

The NaFlex Variant algorithm maintains shared-signal simultaneity across multiple vehicle schedules while observing strict real-time constraints and payload capacities. Its near-linear empirical runtime and slot-count minimization render it practical for large-scale automotive applications. The reliance on First-Fit with period-based ordering achieves slot utilization closely approaching optimal. A plausible implication is that the general methodology may be adaptable to other time-triggered network scheduling contexts where simultaneous signal placements are required.

7. Contextual Significance Within Automotive Systems

The deterministic alignment of signal transmission in FlexRay static segments across multiple vehicle variants increases reliability and serviceability, especially in large-scale platforms reusing hardware and diagnostic tools. This unified multi-schedule formalism supports modern automotive workflows by simplifying signal traceability, facilitating diagnostics, and promoting component reuse, aligning with manufacturing requirements for platform-based design.

For further technical details and benchmark specifications, refer to "Multi-Variant Time Constrained FlexRay Static Segment Scheduling" (Dvořák et al., 2020).