Sensing-Communication Enhanced Control

• Sensing-and-Communication-Enhanced Control is an integrated approach that co-designs sensing, communication, and control for optimized performance in cyber-physical systems.
• It employs real-time, joint optimization of sensor activations, communication resources, and control actions to meet tight latency, energy, and accuracy constraints.
• Advanced algorithms using deep reinforcement learning and Bayesian planning enable robust, adaptive performance in applications like UAV swarms, industrial wireless control, and satellite-UAV networks.

Sensing-and-Communication-Enhanced Control (SCEC) refers to the explicit co-design and real-time interleaving of sensing and communication mechanisms with control logic, transcending traditional modular architectures in which these functionalities are designed in isolation. In SCEC schemes, feedback measurements, sensor activations, communication resources, and control actions are dynamically orchestrated—often in a closed-loop—such that advances or perturbations in one domain (sensing, communication) directly modulate the others for optimal performance under application-specific metrics and constraints. SCEC is foundational in contemporary and emerging cyber-physical networks (e.g., UAV swarms, IIoT, cooperative robotics), enabling reliable, efficient, and robust control at scale under uncertain and dynamic conditions.

1. Principles and System Architectures

SCEC systems deploy highly coupled architectures wherein sensor observations, communication events, and control actuation are tightly integrated across both workflow and design layers. Examples include:

• Integrated Sensing-Communication-Control (ISCC) Frameworks: As instantiated in UAV swarms, the ISCC paradigm closes the loop through sequential operations: a UAV acquires ISAC observations, exchanges communication packets for discrimination between obstacles and peer agents, fuses all available information (including peer-to-peer shared predictions) using an EKF, and initiates control updates based on fused state estimates and collision risk analysis (Wei et al., 21 Jan 2026).
• DS3C Fabric for 6G: This architecture maps four service-level strata—sensing, network (communication), computing, and control—onto an end-to-end loop whereby raw measurements are timestamped, networked (with deterministic slices and URLLC bearers), processed by AI-driven resource managers, and used by digital- or model-based controllers, which in turn actuate physical devices. Slice management, resource orchestration, and controller placement are jointly optimized to meet latency and reliability specifications vital for control stability (Vukobratović et al., 2023).

These architectures are realized via real-time scheduling, dynamic resource allocation, and (in advanced frameworks) distributed learning methods that explicitly account for the interdependencies between sensing, communication, and control.

2. Co-Optimization Methodologies

A defining feature of SCEC is the joint or coupled optimization of system variables cutting across all domains. Methodologies include:

• Implicit and Explicit Co-Design: While some works focus on implicit coupling via adaptive feedback—e.g., control-aware semantic quantization rates or sensor activation (Zhao et al., 8 Oct 2025, Li et al., 6 May 2025)—others formalize joint optimization problems. Typical variable sets include sensor quantization levels, sampling intervals, transmission powers, link scheduling, communication code rate, batch sizes, and location/trajectory variables for mobile agents (Jin et al., 26 Jun 2025, Meng et al., 2024, Meng et al., 2023).
• Optimization Problem Examples:
  • Multi-objective formulations balancing LQR control cost and Fisher information for localization, subject to resource, collision-avoidance, and finite-blocklength (FBL) rate constraints. The optimal LQR cost is linked explicitly to the cumulative communication throughput, and the Fisher matrix reflects the joint sensor placement and transmission power (Jin et al., 26 Jun 2025).
  • Energy efficiency maximization in satellite-UAV networks featuring joint DQN-based trajectory selection, closed-form power control, and interleaved sensing scheduling under remote estimation stability constraints (Liang et al., 2024).
  • Model-predictive-control-inspired receding-horizon problems, where at each control cycle, resource allocation (bandwidth, error probability), feedback gains, and sampling rates are optimized jointly to minimize predicted control cost while satisfying physical-layer queueing and Lyapunov-stability constraints (Meng et al., 2023, Meng et al., 2024).
• Analytical Coupling Inequalities: Several frameworks derive explicit inequalities tying together sensing fidelity, communication reliability, and closed-loop convergence via Lyapunov methods. For example, bounds on the convergence rate $\rho$ of the closed-loop state in terms of quantization level $r$, communication-derived packet-loss $\epsilon_c$, and control gain $K$ (Meng et al., 2024, Meng et al., 2023).

3. Algorithmic Innovations

SCEC has driven a diverse array of algorithmic advances, often leveraging cutting-edge optimization, learning, and estimation techniques:

• Hybrid Semantic–Control Codes: Time-sequence-based semantic communication leverages mutual information estimation (MINE) at the transmitter to avoid sending temporally correlated (redundant) control commands, LSTM-based sequence predictors at the receiver to reconstruct missing signals, and control-gain tuning via reinforcement learning—all integrated in a multi-agent DRL loop optimizing for both communication duty cycle and control error (Li et al., 6 May 2025).
• Self-Triggered and Sparse Control: Self-triggered policies jointly optimize feedback gain sparsity and inter-execution intervals (sampling times), solving mixed $\ell_0$/$\ell_1$/QCQP problems to minimize total sensing, communication, and actuation events under explicit quadratic cost and stability constraints. Time averages show $38\%$ reduction in executed feedback channels and $50\%$ in actuation events for $<15\%$ performance loss (Bahavarnia et al., 2018).
• Active Inference and Bayesian Planning: In UAV SCC systems, an AIF approach formulates a unified generative probabilistic model, executing variational message-passing for state estimation (Kalman filtering) and expected-free-energy minimization for jointly planned sensing-resource allocation and control actions. The result is dynamically balanced sensing–control tradeoffs, with explicit priors reflecting energy and estimation cost, and efficient convergence to joint optima (Pan et al., 17 Sep 2025).
• Alternating Optimization and DRL for Resources: Joint resource allocations in SCEC, such as positioning of movable antennas, beamforming, sensor power, and link schedule, are commonly solved via alternating optimization, particle swarm approaches, projected gradient descent, and, in some settings, deep reinforcement learning for sequential decision processes under real-time feedback (Wang et al., 11 Aug 2025, Liang et al., 2024).

4. Performance Metrics and Empirical Results

Quantitative evaluation of SCEC systems involves benchmarks for control, communication, and sensing that reflect their tightly coupled design:

• Control Performance: LQR cost, cumulative control error, convergence rates, and stability margins. SCEC schemes have demonstrated $20$–$30\%$ reductions in collision avoidance path-planning delay in UAV swarms (Wei et al., 21 Jan 2026), and sub-$2$cm steady-state errors in cooperative mobile robotics (Wang et al., 2023).
• Communication and Sensing Efficiency: Reductions in communication energy (up to $77\%$), number of samples collected ($52\%$), and transmission duty cycles ($85\%$ saving), while matching or exceeding baselines in terminal accuracy and generalization (Cai et al., 14 Feb 2025, Li et al., 6 May 2025). In satellite-UAV and multi-UAV control networks, sensing-communication-enhanced control yields higher energy efficiency and enables operation under tighter power or delay budgets without destabilizing the closed loop (Liang et al., 2024, Jin et al., 26 Jun 2025).
• Robustness and Adaptivity: SCEC methods exhibit significant robustness to noise, packet loss, and non-idealities. For example, semantic communication with control-aware adaptation achieves $5$–$10$ dB PSNR gain under deep-space channel fading and maintains $>90\%$ landing success rate at SNRs where classic JPEG-based pipelines fail (<$70\%$) (Zhao et al., 8 Oct 2025).
• Resource–Performance Tradeoffs: All frameworks highlight sharp transitions: control performance typically improves with more aggressive sensing/communication, but saturates or destabilizes past specific resource thresholds, confirming the necessity of coupled resource–control design (Meng et al., 2023, Meng et al., 2024, Jin et al., 26 Jun 2025).

5. Application Domains

SCEC is foundational in several advanced application settings:

• Autonomous UAV Swarms: Real-time collision avoidance, cooperative reconnaissance, and agile formation maneuvers depend critically on integrating ISAC measurements and low-latency inter-agent communications with control loops (Wei et al., 21 Jan 2026, Jin et al., 26 Jun 2025).
• Industrial Wireless Control Systems (IIoT): Ultra-reliable teleoperation with minimal communication resource usage is achieved using learning-based semantic encoding and transmission-aware adaptive control, affecting industrial robot arms, automated guided vehicles, and remote haptics (Li et al., 6 May 2025, Meng et al., 2023).
• Satellite-UAV-Edge Architectures: Remote UAV control and IoT data aggregation beyond terrestrial coverage require careful balancing of trajectory, sensing, and power to sustain stable energy-efficient operation under satellite bandwidth and link constraints (Liang et al., 2024).
• Cooperative Robotics with Optical ISAC: Where radio is unavailable, joint use of optical camera sensing and screen-camera communication (SCC) enables both pose determination and command sharing, enabling responsive, robust mobile formations (Wang et al., 2023).
• 6G and Tactile Internet: DS3C fabric natively integrates control objectives with low-latency networking, enabling closed-loop services such as distributed haptic interfaces, remote surgery, and collaborative cobotics at sub-millisecond jitter (Vukobratović et al., 2023).

6. Challenges and Future Research Directions

Key technical challenges in SCEC include scalable distributed optimization under resource constraints, real-time learning for nonlinear or partially observed environments, incorporating quantization and coding limits, robustness against network adversities (delay, outage), and cross-layer security. Further open directions are:

• Advanced resource orchestration in large-scale multi-tenant and multi-hop networks, especially with mobile or dynamically reconfigurable agents (e.g., movable antennas, UAV relays) (Wang et al., 11 Aug 2025).
• Theoretical analysis beyond quadratic or linear-Gaussian regimes: tight fundamental limits for control under semantic communication, batch learning, or non-iid stochastic processes.
• Integration of physical-layer innovations (e.g., reconfigurable intelligent surfaces, ISAC transceivers) with network- and application-layer policies for holistic, adaptive control (Vukobratović et al., 2023).
• Deep reinforcement learning and active inference at scale for online adaptation to unforeseen physical or network conditions.
• Practical zero-touch automation and explainable AI-in-the-loop for trustworthy SCEC deployment in safety-critical domains.

The SCEC paradigm is thus central to next-generation cyber-physical systems, demonstrating that only by orchestrating sensing, communication, and control as inseparable, feedback-driven entities can one realize robust, efficient, and responsive control under uncertain, dynamic, and resource-constrained regimes.