Integrated Communication-Sensing-Control

• Integrated Communication-Sensing-Control is a systemic paradigm where sensing, communication, and control merge into a real-time closed loop.
• It jointly optimizes resource allocation, control policies, and information flow to boost performance in cyber-physical systems, UAV swarms, and industrial automation.
• Research in ICSC explores threshold-based switching, optimal control techniques, and neural co-design, offering actionable guidelines for next-generation systems.

Integrated communication-sensing-control (ICSC), also referred to as integrated sensing, communication, and control (ISCC), denotes a systemic paradigm in which the traditional separation of sensing, communication, and control is replaced by a deeply coupled architecture, forming a real-time closed feedback loop. In ICSC/ISCC architectures, information from sensors is fused, transmitted, and fed back to actuators or controllers with joint optimization of resource allocation, control policy, and information flow. This co-design enables new levels of performance, robustness, efficiency, and scalability, particularly in cyber-physical systems, networked robotics, wireless industrial automation, and emerging 6G networks.

1. Fundamental Concepts and Architectural Models

ICSC frameworks consist of three tightly intertwined subsystems:

• Sensing: Acquisition of state estimates, environmental parameters, or process measurements through local or distributed sensor arrays (e.g., radar, LiDAR, cameras, inertial units) (Wei et al., 21 Jan 2026). Modern designs increasingly exploit multi-modal or cooperative sensing for redundancy, coverage, and accuracy.
• Communication: Reliable, low-latency, and often wireless exchange of control-relevant information, using shared channels that are also opportunistically used for environmental sensing (i.e., ISAC principles) (Soleymani et al., 30 Jan 2026, Li et al., 31 Mar 2026).
• Control: Real-time computation of actuation or trajectory commands, fed back to remote devices with tight performance guarantees on stability, energy, or tracking error.

These subsystems are coupled via a feedback dataflow: sensor observations are processed, encoded, transmitted (or reconstructed if transmissions are not initiated), and then used to update control or estimation policies in real time (Li et al., 6 May 2025). This feedback loop is then closed by actuation and (optionally) adaptive reconfiguration of sensors and communication resources.

Mathematical co-optimization formulations often take the form:

$$\max_{\{u_k, \mathbf{w}_k\}} \sum_{k=0}^K \left[ \alpha_s \hat{s}_k^s(\mathbf{w}_k) + \alpha_c R_k(\mathbf{w}_k) - \alpha_\ell L_k(\mathbf{w}_k) - \alpha_e e_k(u_k) \right]$$

subject to process dynamics, stability, sensing/communication/control quality constraints, and real-time resource budgets (Wei et al., 21 Jan 2026).

2. Joint Resource Allocation and Performance Coupling

ICSC frameworks reveal strong interdependencies among the subsystems. Resource allocation (power, bandwidth, time-frequency resources, quantization, spatial degrees-of-freedom) directly impacts the achievable estimation error, communication reliability, and closed-loop control performance:

• Sensing quantization level $r$ determines the uplink rate and estimation error variance (Meng et al., 2024).
• Communication capacity (often under queueing, fading, or finite-blocklength constraints) governs packet loss, latency, and effective downlink control delivery (Meng et al., 2024, Li et al., 31 Mar 2026).
• The design parameters (sampling, sensing rate, communication bandwidth/power, and feedback gain) must be jointly tuned to guarantee prescribed convergence or stability rates (cf. Lyapunov or Riccati-based conditions) (Meng et al., 2024, Soleymani et al., 30 Jan 2026).

Key theoretical results include:

• Analytical performance inequalities relating estimation error, communication-induced loss/delay, quantization, and convergence rate $\rho$ (Meng et al., 2024):

$$\rho \geq \frac{F_\rho(W_u, W_d, K, \mathbf{X}_t)}{\|\mathbf{X}_t\|_P^2}$$

• Explicit closed-form thresholds for mean-square stability of discrete-time stochastic systems under packet drops and control signaling failures (Li et al., 31 Mar 2026):

$$\rho(M) < 1 \iff \alpha_\mathrm{ctrl} > \alpha_\mathrm{ctrl}^*$$

These performance coupling laws serve as design guidelines, demonstrating that sensing, communication, and control processes are inseparable: deficiency in any one will bottleneck the overall system.

3. Optimal Control and Scheduling Policies in ICSC

Optimal control policies are fundamentally altered by information constraints and uncertainty in communication and sensing channels. Key developments include:

• Threshold-based Switching: In linear-quadratic-Gaussian (LQG) settings with unreliable links, the optimal mode (whether to sense or communicate at each step) is defined by a matrix-valued threshold on estimation covariances (Soleymani et al., 30 Jan 2026):

$$u_t^* = \begin{cases} 1 \text{ (communicate)}, & \text{if } P_t \succeq T_t(Q_t) \ 0 \text{ (sense)}, & \text{otherwise} \end{cases}$$

The switching region shrinks as base-station uncertainty increases and expands with source/process uncertainty.

• Certainty-Equivalent Control: When state estimates are available, the optimal controller is typically linear in the estimated state, with gains calculated via Riccati recursion contingent on real-time communication and estimation statistics (Soleymani et al., 30 Jan 2026, Chen et al., 5 Feb 2026).
• Stochastic MPC with Communication-Sensing Constraints: The closed-loop design can be cast as a convex quadratically-constrained quadratic program (QCQP) accounting for future constraints on rate, SNR, and communication-induced uncertainty, often recursively solvable in real time (Chen et al., 5 Feb 2026).

In architectures with dynamically shareable resources, joint metaheuristic or alternating optimization algorithms (e.g., differential evolution (Meng et al., 2024), successive convex approximation (Li et al., 31 Mar 2026), or alternating optimization with metaheuristics and convex relaxation (Wang et al., 11 Aug 2025)) are adopted to handle the intrinsic non-convexity.

4. Physical and System Layer Integration

ICSC systems are implemented across a range of platforms and physical layers:

• Wireless UAV Swarms: Multirotor or fixed-wing agents equipped with multimodal sensors, distributed controllers, and software-defined radios for wireless ISCC (Wei et al., 21 Jan 2026, Chen et al., 5 Feb 2026, Wei et al., 11 Feb 2025). Over-the-air (OTA) aggregation allows for spectrum-efficient collection and broadcast of control/sensing data in large-scale swarms.
• Industrial Automation: Closed-loop process control networks, where quantization, queueing, and wireless link unreliability are explicitly modeled for tight control of actuators via remote controllers (Meng et al., 2024).
• Adaptive Lighting/VLC: Smart building and 6G systems leveraging integrated optical transceivers for joint data and illumination delivery, with non-line-of-sight (NLOS) sensing and dynamic mode switching for energy-aware adaptation (Xie et al., 26 Nov 2025).
• Mobility and Intelligent Surfaces: Physical relocation of antenna arrays for dynamic beamforming and QoS guarantees to distributed users, targets, and controlled vehicles in joint communication, sensing, and control tasks (Wang et al., 11 Aug 2025).

Resource-physical coupling is exemplified by the direct mapping between time-frequency/antenna resource fractions allocated for sensing, control, and communication and the resulting closed-loop mean-square stability and trajectory-tracking accuracy (Li et al., 31 Mar 2026).

5. Algorithmic and Neural Approaches: Semantic and Adaptive Co-Design

Modern ICSC systems increasingly employ learning-based co-design for semantic communications and control:

• Semantic Communication and Control: Adaptive transmission and inference policies are formed using mutual information estimation, LSTM-based semantic feature reconstruction, and actor-critic RL for dynamic gain adaptation (Li et al., 6 May 2025). Update policies suppress redundant communication, reconstruct unseen data, and adapt control based on instantaneous semantic value and communication channel quality, achieving ultra-low duty cycles while maintaining tracking accuracy.
• Hybrid Reward Multi-Agent RL: Multi-agent (update/gain) architectures are trained via hybrid rewards representing control deviation and communication usage, surpassing classical periodic or always-on schemes in duty-cycle reduction and real-world teleoperation performance (Li et al., 6 May 2025).

6. Performance Gains, Metrics, and Design Insights

Comprehensive system-level studies reveal substantial performance improvements:

• LQG Cost Reduction: Integrated design can improve LQG cost by 20–30% over separate (non-joint) heuristics via uncertainty-aware scheduling (Soleymani et al., 30 Jan 2026).
• Tracking Accuracy: ISCC frameworks for UAVs yield decimeter-level trajectory tracking given channel losses and finite blocklength, halving or bettering the error compared to GNSS or naive design baselines (Li et al., 31 Mar 2026, Chen et al., 5 Feb 2026).
• Duty Cycle and Energy: Semantic ISCC architectures reduce wireless duty cycle by over 85% compared to always-on control, with superior accuracy (Li et al., 6 May 2025).
• Spectrum and Power: OTA-based ISCC in UAV swarms allows N-fold spectrum reuse compared to per-agent allocation, achieving near-optimal control and sensing performance (Wei et al., 11 Feb 2025). Coordinated power control reduces transmit power while meeting tightly coupled SINR-CRLB tradeoffs in distributed ISAC systems (Huang et al., 2022).

Metrics quantifying ISCC performance include MMSE/CRLB for estimation, data rate, latency, packet success probability, stability margin (maximum spectral radius of closed-loop system), tracking error, and application-specific measures such as illuminance or SNR uniformity (Wei et al., 21 Jan 2026, Xie et al., 26 Nov 2025, Huang et al., 2022).

7. Challenges, Extensions, and Future Directions

Critical ongoing research directions include:

• Extension of joint design to wideband MIMO, non-linear or non-Gaussian systems, actuator and antenna mobility, and multi-agent cooperative tasks.
• Integration of end-to-end neural architectures mapping raw sensor data and traffic demands to control and communication actions (Wei et al., 21 Jan 2026).
• Exploitation of quantum-enhanced sensing, cross-layer security strategies, and the development of digital twins for off-line optimization and anomaly-aware control (Wei et al., 21 Jan 2026).
• Robustification against time-varying channels, actuator saturation, and adversarial uncertainty via robust optimization and online learning.

The convergence of sensing, communication, and control into fully integrated, computationally tractable, real-time feedback systems is now foundational for next-generation cyber-physical, industrial IoT, and autonomous networked system design (Soleymani et al., 30 Jan 2026, Wei et al., 21 Jan 2026, Li et al., 31 Mar 2026, Li et al., 6 May 2025).