Adaptive AI-Driven Waveform Design

This Army SBIR solicitation (OSW26BZ04-DV009) seeks to develop an AI/ML-controlled radar waveform generator that dynamically adjusts frequency, modulation, and coding in real time. The goal is to optimize radar performance and evade advanced jamming, digital radio frequency memory (DRFM) systems, and radio frequency interference (RFI) in contested environments. The project leverages deep reinforcement learning (DRL) to enable a neural agent to autonomously synthesize and optimize transmit waveforms based on real-time spectral observations. Because this capability requires micro-second execution, the models must be optimized for edge-computing architectures, such as software-defined radios (SDRs) and field-programmable gate arrays (FPGAs), while adhering to strict size, weight, and power (SWaP) constraints. The project is structured into three phases: - **Phase I:** Conduct a feasibility study and develop a simulated software proof-of-concept demonstrating improved target detection over traditional fixed-waveform baselines. - **Phase II:** Implement and optimize the AI/ML controller onto physical, edge-computing hardware (SDR/FPGA) and deliver a fully integrated prototype for laboratory or field testing. - **Phase III:** Transition the technology into military platforms, such as UAS sensors, ground surveillance, and AESA radars. This technology has dual-use potential for commercial automotive radar, air-traffic control, and dynamic spectrum sharing. The topic is restricted under ITAR/EAR regulations, requires a CMMC Level 2 (Self) certification, and falls under the "Trusted AI and Autonomy" and "Integrated Sensing and Cyber" modernization priorities.