<p>Unmanned Aerial Vehicles (UAVs) are rapidly deployed for sustainable remote sensing applications such as environment monitoring, disaster monitoring, infrastructure inspection, etc. For such real-time awareness and decision-making scenario, simultaneous data sensing and transmission plays a crucial role. Joint Radar and Communication (JRC) waveforms address this requirement and also enhances spectrum usage and operational capabilities in UAV systems. However, the highly dynamic UAV conditions, including changes in height, velocity, and signal interference, make precise classification of JRC waveforms a significant challenge. In this work, a framework is proposed for JRC waveform classification with Cramer Rao Lower Bound (CRLB) Regularized Deep Adaptive Learning, which employs hybrid Attention Convolutional Neural Network (CNN) and Harmony Search (HS) algorithm guided by CRLB for uncertainty aware adaptation. For capturing oscillatory characteristics, Morlet Wavelet Transform is employed aiming time-frequency decomposition and multi-scale characteristics essential for precise classification. The framework integrates multi-scale and multi-angle analysis, accounting for angular variations such as orientation and motion dynamics, ensuring reliable performance across diverse UAV scenarios. Various constraints, including root mean square error, pitch adjustment rate, bandwidth, classification accuracy, etc., are considered to evaluate the effectiveness of the proposed algorithm in UAV applications. The proposed framework not only improves waveform classification accuracy under varying conditions for both experimental collected and synthetic dataset, but also pave a path towards sustainable remote sensing.</p>