Integrated Sensing, Communication & Control

• Integrated Sensing, Communication and Control is a unified paradigm that combines environmental sensing, data exchange, and actuation into a closed-loop system.
• The framework employs methods like active inference, alternating optimization, and semantic DRL to co-optimize resource use and enhance system performance.
• ISCC enables advanced applications in UAV swarms, industrial IoT, and 6G networks by reducing latency, improving tracking accuracy, and boosting throughput.

Integrated Sensing, Communication and Control (ISCC) unifies the traditionally siloed functions of environmental sensing, data exchange, and decision/actuation within a tightly coupled, systemic paradigm. Unlike isolated designs, where each module is optimized independently, ISCC establishes a closed loop in which information flow is explicitly coordinated and co-optimized, typically yielding significant performance benefits in terms of latency, efficiency, robustness, and resource utilization. The ISCC paradigm is central to emerging applications in UAV swarms, industrial automation, robotic systems, and future 6G low-altitude wireless networks.

1. Theoretical Foundations and Problem Formulations

ISCC systems jointly optimize sensing, communication, and control variables under a unified mathematical or algorithmic framework. Particularly, recent formulations explicitly encode the cross-dependence between resource allocation, estimation, and action selection:

• Active Inference Framework (AIF): Pan et al. (Pan et al., 17 Sep 2025) model the ISCC problem for UAVs as minimizing the sum of variational free energy (VFE) for state estimation and expected free energy (EFE) for action planning within a fully generative Bayesian model:

$$P_g(x_{1:T},y_{1:T},u_{1:T-1},k_{2:T}) = P_{g,1}(x_{1:t},y_{1:t}|u_{1:t-1},k_{1:t}) P_{g,2}(x_{t+1:T},y_{t+1:T}|x_t,u_{t:T-1},k_{t+1:T}) \prod_{\tau=t}^{T-1}\tilde P(x_{\tau+1},y_{\tau+1},u_\tau,k_{\tau+1})$$

Joint inference and planning are achieved by minimizing $F[q]$ (VFE) over posterior beliefs and $G[a]$ (EFE) over control and communication actions.

• Finite-Horizon Co-Optimization: In ISCC-enabled UAV swarms, system state $x_k$, control $u_k$, local sensing $y_k^s$, and communication measurements $y_k^c$ are jointly optimized:

$$\min_{u_{0:N-1},\theta_{0:N-1}} J_\text{total} = \alpha J_\text{sensing} + \beta J_\text{comm} + \gamma J_\text{ctrl}$$

Subject to control limits, communication and sensing budgets, estimation error, latency, and safety constraints (Wei et al., 21 Jan 2026).

• Semantic ISCC: ISC³ extends ISCC by embedding semantic mutual information criteria and control-oriented feature abstraction in real-time robot communication (Li et al., 6 May 2025).

2. Architectures and Enabling Mechanisms

ISCC frameworks exhibit modular or layered architectures with deeply coupled data paths and cross-domain information reuse:

• Module Integration: Each agent (e.g., UAV or robot) integrates digital signal processing for sensing, communication PHY/MAC, and control logic via shared memory or data buses, favoring rapid cross-module propagation of state estimates, control commands, and channel/environmental context (Wei et al., 21 Jan 2026).
• Shared RF Front-Ends: Utilization of common RF hardware for both radar and communication supports simultaneous environment mapping and data exchange.
• Swarm-Level Coordination: Information sensed locally is disseminated throughout the agent collective, enabling global state awareness and formation control.
• Movable Antenna Arrays: By endowing BSs with physically reconfigurable antenna arrays (MAs), the system adaptively shapes the RF aperture to optimize joint communication, sensing, and control objectives (Wang et al., 11 Aug 2025).

3. Algorithmic and Optimization Methods

ISCC optimization is characteristically high-dimensional and coupled, demanding advanced algorithmic strategies:

Paper	Optimization Technique	Resource Variables
(Pan et al., 17 Sep 2025)	Active inference (AIF), factor-graph message passing	Control $u_t$, subcarrier allocation $k_t$
(Wang et al., 11 Aug 2025)	Alternating optimization (PSO + convex SDP)	Antenna positions $\mathbf{p}$, beamformers $\mathbf{w}$
(Wei et al., 21 Jan 2026)	Alternating minimization, MPC, SOCP	Control $u_k$, comm allocation $\theta_k$
(Li et al., 6 May 2025)	HR-MADRL, semantic feature learning	Tx activation $a_i$, control gain $K_i$

• AIF for ISCC: Factorizes the system dynamics, observation, and resource allocation into single generative models, leading to efficient closed-loop planning via variational inference and message passing (Pan et al., 17 Sep 2025).
• Alternating Optimization (AO): Decouples position and beamforming optimization in non-convex ISCC problems, using metaheuristics (e.g., PSO) for geometry/topology and convex solvers for beam and covariance design (Wang et al., 11 Aug 2025).
• Distributed MPC: Embeds sensing and communication quality terms in the classical MPC cost function. Kalman filters, SOCP, and QP solvers alternate to co-optimize tracking, estimation, and communication resource allocation (Wei et al., 21 Jan 2026).
• Semantic DRL: Employs mutual-information neural estimation and multi-agent RL to regulate semantic transmission probability and adaptive control gains, achieving major reductions in communication load (Li et al., 6 May 2025).

4. Resource Coupling, Trade-offs, and Performance Metrics

ISCC fundamentally involves managing trade-offs among estimation accuracy, communication throughput, control energy, and latency.

• Resource–Performance Coupling: Control- and sensitivity-aware resource allocation aligns communication and sensing bandwidth to instantaneous estimation uncertainty and control objective tightness.
• Cost Integration: Unified cost/prior formulations (e.g., AIF's $\tilde P$) encode trade-offs as in $J_t^{control}(x,u)$ and $J_t^{sensing}(k)$ (Pan et al., 17 Sep 2025).
• Empirical Performance:
  • Pan et al.: Proposed AIF reduces both control ($3.74\times10^3$) and sensing cost ($-5.23\times10^3$) compared to greedy and random baselines (Table I in (Pan et al., 17 Sep 2025)).
  • MA-based ISCC achieves 15–30% communication throughput gains vs. static arrays, with comm–control–sensing constraints satisfied (Section 5.2 of (Wang et al., 11 Aug 2025)).
  • ISC³ demonstrates an 85% reduction in communication duty cycle at sub-3 mm tracking error (Figs. 5–6 in (Li et al., 6 May 2025)).
  • ISCC-based UAV swarms reduce trajectory tracking error from ∼10 m (GNSS only) to 0.22 m (Section 2, (Ma et al., 8 Dec 2025)).

5. Representative Applications and Case Studies

ISCC has been instantiated in diverse multi-agent robotics and wireless control scenarios.

• UAV Swarms: Core applications include disaster relief, aerial BSs, and logistics (Wei et al., 21 Jan 2026), featuring:
  • Real-time blank-band OFDM echo recovery via signal/model fusion.
  • Kalman-filter-based multi-view data fusion and event-triggered information exchange.
  • Predictive beamforming using radar-augmented channel state priors.
  • Sensing-informed neighbor discovery and adaptive routing frameworks.
• Low-Altitude Wireless Networks: ISCCC enables trajectory-optimized, safe, and low-latency flying in LAWN ecosystems via layered architectures that merge S, C, C', and K at flying, functioning, and self-organizing planes (Ma et al., 8 Dec 2025).
• Semantic Control for Industrial IoT: Time-sequence ISCC with mutual-information-based sparsification sharply reduces resource usage in teleoperation and control tasks (Li et al., 6 May 2025).

6. Limitations, Challenges, and Research Directions

Current ISCC realizations face open scientific and engineering challenges:

• Scalability: Scaling to dense, heterogeneous swarms strains S/C resources; proposed mitigations include next-generation multiple access, spatio-temporal semantic communication, and distributed computing (Ma et al., 8 Dec 2025).
• Real-Time Constraints: Control-loop stability, finite blocklength, and end-to-end delay require joint optimization of power, computation, and comm scheduling.
• Robustness: Adversarial jamming, hardware mismatches, and channel uncertainty have yet to be comprehensively incorporated in deployed ISCC loops (Wei et al., 21 Jan 2026).
• System Modeling: ISCC control generally leverages linear-Gaussian system models; extensions to nonlinear, non-Gaussian, or mixed-integer regimes (e.g., via extended/unscented AIF or hierarchical planning) are noted as future directions (Pan et al., 17 Sep 2025).
• Architectural Innovations: Emerging research explores virtual distributed MIMO arrays via swarm formation control and reconfigurable AI-driven aerial infrastructure (Ma et al., 8 Dec 2025).
• Waveform and Protocol Co-Design: AI-generated ISAC waveforms and semantic-aware protocol stacks tuned to control–sensing–comm requirements are active areas (Ma et al., 8 Dec 2025).

7. Summary Table: Key ISCC Works and Technical Contributions

Reference	Domain/Application	Key Techniques/Findings
(Pan et al., 17 Sep 2025)	UAV single-agent ISCC	Active inference, unified generative model, Bayesian control
(Wang et al., 11 Aug 2025)	BS–UAV with movable antennas	Alternating optimization, beamforming, MA positioning
(Wei et al., 21 Jan 2026)	UAV swarm ISCC	Real-time co-optimization, MPC, distributed estimation
(Ma et al., 8 Dec 2025)	ISCCC in LAWN for UAV swarms	Three-layer architecture, joint trajectory optimization
(Li et al., 6 May 2025)	Time-sequence ISC³ for semantic control	Mutual-information estimation, HR-MADRL, LSTM reconstruction

Each of these works demonstrates how tightly integrating sensing, communication, (computing,) and control produces measurable gains over independent module approaches, particularly in the presence of communication constraints, dynamic environments, and strict performance requirements.