- The paper introduces task-aware AoA and CoMA to quantify and optimize actuation timeliness and reliability in WNCS.
- It models WNCS using Geo/D/C/C and Geo/Geo/C/C queues to compare deterministic and stochastic service regimes.
- Differentiated resource allocation is shown to be essential for balancing network contention and meeting heterogeneous QoS constraints.
Semantics-Aware Communication and Differentiated Resource Allocation in WNCS
This paper addresses the optimization of timeliness and reliability in Wireless Networked Control Systems (WNCS) through a semantics-aware joint communication-computation perspective. The core concept is that sampled data streams are categorized as regular or critical based on their semantic states, driving differentiated resource allocation. Semantics-aware resource management is formalized by introducing two task-centric metrics: Task-aware Age of Actuation (AoA), a refinement on AoI focusing strictly on actuation freshness, and the Cost of Missing Actuation (CoMA), a weighted penalty reflecting the impact of failed actuations. The WNCS is structurally modeled as a sensor-controller-actuator cascade, with resource contention and actuations managed by a multi-rate Geo/D/C/C queue under finite computation resources.
The uplink is characterized by Nakagami-m fading and large-scale path loss, with transmission power allocations tightly bounded by device energy budgets. The server pool at the controller processes tasks based on semantic class: simple regular tasks require single-threaded computation, whereas complex critical events are given parallelized compute allocations. Admission and resource-allocation policies are semantics-driven and non-preemptive, leading to loss systems with immediate task rejection in case of insufficient resources.
Traditional AoI fails to capture the semantic utility of information as it ignores the nuances of actuation and task heterogeneity. The key contributions here are:
- Task-aware AoA: This metric quantifies the time gap since the generation of the most recently executed command for each task class, emphasizing execution, not just reception, timeliness. Exact closed-form expressions for time-average AoA are derived given stochastic arrivals, service, and non-preemptive queueing.
- CoMA: Generalizing beyond average drop rate, CoMA encodes differentiated penalties (e.g., catastrophic for critical tasks, negligible for regular ones) and is analytically derived as a function of arrival rates, wireless transmission success, and probability of computational admission.
The joint use of AoA and CoMA enables accurate Pareto optimization in semantics-aware WNCS, balancing timeliness of regular control with robust reliability guarantees for high-stakes critical events.
Queueing Analysis and Resource Dynamics
The queueing subsystem is tackled using two regimes:
- Deterministic Service (Geo/D/C/C): System state is a composite of the remaining execution times across active tasks (exact dimensionality explodes combinatorially in resource units and service time). State transitions are modeled as deterministic shift operations, and steady state is computed using balance equations with normalization.
- Randomized (Geometric) Service (Geo/Geo/C/C): By sacrificing deterministic tracking for memoryless geometric service, state-tracking complexity drops to two dimensions (active counts of regular and critical tasks). Transitions and admissions are calculated via binomial and indicator-driven probability mass functions, and closed-form steady-state blocking probabilities are derived. This regime closely approximates the deterministic one in the light-traffic scenario.
- Erlang’s Multirate Loss Model: For further analytical tractability, blocking and occupancy probabilities are upper-bounded using the M/G/C/C model (Erlang) and shown to provide tight bounds except at high loads.
Strong numerical evidence is presented: deterministic queues give the lowest blocking probability; stochastic and continuous-time models offer effective upper bounds and analytic tractability with minimal loss of fidelity for moderate to light loads.
System objectives are cast into a bi-objective optimization framework: minimize (i) long-term CoMA and (ii) regular task AoA, subject to energy/arrival constraints and physical limits on the differential power and admission probabilities.
The numerical analysis demonstrates:
- Resource contention causes regular and critical streams to interfere: increasing admission probability for regular tasks degrades both the AoA of critical tasks and system-level CoMA.
- Differentiated allocation across semantic streams is essential for system efficiency: sacrificing regular tasks in resource-starved conditions is optimal to maximize critical-task reliability.
- The approach enables strict timeliness-constrained actuation for regular flows, while ensuring ultra-reliability for critical scenarios, leveraging the inherent error tolerance of regular streams.
- Task-aggregated AoI offers limited insight in contrasting trade-offs, while task-aware AoA and CoMA offer transparent and actionable optimization levers.
Simulation results consistently validate the analytic models. The system’s Pareto frontier, derived over resource allocation policies, displays clear trade-offs between AoA and CoMA, revealing differentiated strategies’ clear superiority compared to undifferentiated baselines.
Theoretical and Practical Implications
This work provides a robust compositional framework for semantics-aware communication and computation co-design in WNCS under strict resource constraints. By integrating semantic-awareness directly into both metrics and resource scheduling mechanisms, it rigorously demonstrates that differentiated allocation policies are both necessary and sufficient for meeting heterogeneous QoS (timeliness, reliability) requirements in real-time multi-task control. The analytic reduction to multirate loss systems, together with resource admission mechanisms, offers a pathway to scalable deployment in complex industrial and safety-critical automation contexts.
The findings also highlight key theoretical insights:
- Departures from memoryless service fundamentally alter state-space dimensionality and thus analytic complexity.
- Upper-bounding exact models using random-service and continuous-time approximations is justified for broad operating regions, facilitating tractable closed-form design.
- Non-preemptive semantics-driven admission is essential for ultra-reliable low-latency (URLLC) regimes.
Future developments may leverage these models for adaptive online scheduling, RL-driven policy synthesis under non-stationary traffic, and further exploration of the control-theoretic stability boundaries under semantics-aware perception and actuation.
Conclusion
This paper delivers a rigorous semantics-aware optimization framework for WNCS, underpinned by exact multi-class multi-server queueing models and validated optimization of actuation freshness and reliability metrics. Differentiated resource allocation, grounded in semantic utility, is analytically and empirically shown to be necessary for the reliable operation of resource-constrained, real-time control systems. The joint modeling and optimization approach sets new standards for integrating communication, computation, and semantics in the design of ultra-reliable, low-latency cyber-physical systems (2605.09095).