Wenwen Wu

Reliable beam management in Unmanned Aerial Vehicle (UAV)-assisted Integrated Sensing and Communication (ISAC) systems needs accurate trajectory prediction and a clear sense of prediction risk. Most existing methods use deterministic future positions or raw uncalibrated uncertainty. Under high mobility and uncertainty, this leads to unreliable beam decisions. We design a control-oriented probabilistic trajectory prediction framework. It uses calibrated trajectory uncertainty as a risk signal for adaptive beam management. The framework first combines motion history and visual context to predict trajectory distributions. Split conformal calibration turns raw Gaussian uncertainty into statistically reliable risk bounds. A codebook-constrained beam management strategy adjusts beamwidth based on the calibrated spatial risk. This balances beamforming gain, coverage robustness, and switching stability. Tests on UAV data show better prediction accuracy than representative probabilistic baselines. The raw uncertainty remains under-calibrated, and conformal calibration is therefore applied to improve its reliability before beam-control decisions. Using the calibrated uncertainty for beam control improves communication coverage and cuts outages and severe misalignment in high-risk situations. Calibrated predictive uncertainty can serve as an actionable control variable for robust beam management in dynamic UAV-assisted ISAC environments.

Unmanned Aerial Vehicles (UAVs) are gradually emerging as key platforms for Integrated Sensing, Communication, and Computation (ISCC) systems in next-generation wireless networks. However, strict resource constraints and task coupling make static allocation inefficient in dynamic environments. This paper studies a UAV-driven ISCC system in which a single UAV dynamically allocates communication bandwidth, sensing resources, and computing power. Considering that sensing data in mission-critical applications is highly time-sensitive, minimizing the response time is paramount. To reduce system latency while maintaining sensing quality and energy efficiency, we propose D3PG-Light, a deployment oriented and stability-enhanced refinement of the deep reinforcement learning framework, specifically tailored for real-time resource scheduling under UAV hardware constraints. D3PG-Light incorporates an adaptive gradient stabilization mechanism, Long Short-Term Memory (LSTM), and feature fusion to enhance training stability. Simulation results based on real air–ground channel measurements show that D3PG-Light converges faster and achieves more stable learning behavior than DDPG, TD3, and the original D3PG. In particular, the proposed method reduces the 95th-percentile latency from over 100 ms to approximately 24 ms, achieves higher converged reward values, and requires fewer than 50 k model parameters. These results demonstrate the effectiveness of D3PG-Light for latency-sensitive UAV-ISCC applications.