AI Based ATR (ABA) Technologies for Radar

(U) Request for Information (RFI): AI Based ATR (ABA) Technologies for Radar (U) Radar Division (U) C5ISR Center, Research & Technology Integration (RTI) Directorate (U) U.S. Army Combat Capabilities Development Command (DEVCOM) 1. (U) Introduction (CUI) The U.S. Army is conducting market research to identify potential Artificial Intelligence/Machine Learning (AI/ML) solutions for Automatic Target Recognition (ATR) using Synthetic Aperture Radar (SAR) imagery. This Request for Information (RFI) is issued to solicit information from qualified vendors regarding their capabilities and approaches to address challenging ATR problems, particularly within the context of the Army's multi-domain Intelligence Surveillance Reconnaissance (ISR) systems. The information received will be used to inform future acquisition strategies and potential program development. This RFI is for informational purposes only and does not constitute a commitment to procure any services or products. 2. (U) Background (CUI) The Army is seeking to enhance its ability to reliably detect, locate, and identify military targets in complex SAR imagery. ATR systems face challenges when dealing with scenarios involving limited training data, challenging environmental conditions, and diverse radar sensors & platforms. This RFI focuses on identifying innovative AI/ML approaches that can overcome these limitations and provide robust ATR capabilities integrated with Army ISR radars. 3. (U) Challenges & Areas of Interest (U) The Army is interested in solutions that address the following challenges: (U) Target / Platform Challenges - (U) Radar Platform Diversity: The Army utilizes a variety of airborne SAR systems. Solutions should minimize the need for extensive re-training and adaptation when transitioning between different radar systems. - (CUI) Under-Represented Targets: ATR performance degrades significantly when dealing with target classes for which only a limited number of real-world SAR samples are available. - (CUI) Occluded Targets: Detecting and recognizing targets that are partially obscured by shadows, terrain, or other objects presents a significant challenge. - (CUI) Challenging Collection Geometries: Army ISR sensing presents a variety of imaging geometries, including long standoff and/or high-altitude applications which cause extreme imaging angles; while potentially providing valuable information, these geometries can introduce complexities that impact ATR performance. (U) Algorithm Challenges - (U) Large Image size: very large images impose memory and/or timeline issues on ATR algorithms. - (U) Synthetic Data Generation & Training: the Army expects to develop ATR solutions for radar systems that are in development. As such, the Army desires techniques for generating high-fidelity synthetic SAR data to augment limited real-world datasets. Emphasis should be placed on evidence-based practices for bridging the gap between synthetic and real data. - (U) Scalable ATR Architectures: The need for integration into multiple platforms presents the challenge and opportunity to scale ATR model size and complexity to the available computational resources and processing capabilities (e.g., edge processing). The Army is interested in understanding the trade-offs between accuracy, speed, and resource requirements. - (U) Hybrid / Mixture of Experts: Scalable architectures may utilize multiple diverse ATR models when adequate processing is available. This capability requires the ability to federate and fuse models to maximize performance. (U) Special Research Challenge (CUI) In addition to SAR ATR for stationary targets, the Army is interested in understanding the technology required to identify movers in SAR. The ability to maintain custody on targets by locating and recovering mover signatures provides the Army with enhanced situational awareness, but the feasibility of this technology is unknown for the current set of Army platforms. Although this RFI mainly addresses ATR for monostatic radar systems, responses may include more advanced collection modes (e.g. distributed). Addressing this research challenge will inform (1) the design and operation of both current and future Army radars and (2) the direction of future ATR development. 4. (U) Information Requested (U) Respondents are requested to provide information in the following areas. Responses should be concise and focused on capabilities relevant to the challenges outlined above. - (U) Brief Company Overview: describe your company and its experience in AI/ML for SAR ATR. - (U) Technology Description: Provide a detailed description of your proposed AI/ML approach, including: o (U) Algorithms and techniques used o (U) Approach to full-scene image handling o (U) Training Data requirements (real and synthetic) o (U) Training methodologies o (U) Performance metrics (e.g., probability of detection, probability of ID, false alarm rate) achieved on relevant datasets. o (U) Algorithm storage requirement - (U) Scalability & Adaptability: Describe how your solution can be scaled to accommodate different hardware platforms and radar systems. - (U) Experience in Integration: Describe your understanding of the ability to integrate ATR on an airborne platform - (U) Technology Readiness Level (TRL): Indicate the current TRL of your technology, including rationale - (U) Potential Challenges & Risks: Identify any potential challenges or risks associated with implementing your solution. - (U) Cost of integration with a new monostatic radar. - (U) Contact Information: Provide contact information for a technical point of contact. 5. (U) Response Format & Submission (U) Responses should be submitted electronically in PDF or MS Word format within 30 days of the RFI release date. Responses should not exceed 10 pages. (U) Submissions must be sent to the following, based on classification: - (U) usarmy.apg.devcom-c5isr.mbx.rti-industry-engagement@army.mil (UNCLASSIFIED) - (U) usarmy.apg.devcom-c5-isr.mbx.rti-industry-engagement@mail.smil.mil (CLASSIFIED) (U) Questions and requests for clarification of information may be submitted no later than 25 calendar days following the RFI posting. (U) NOTE: for CLASSIFIED questions or submissions, please send a courtesy email to the UNCLASSIFIED email address to notify of the request on the Classified side. 6. (U) Disclaimer (U) This RFI is for informational purposes only and does not constitute a commitment to procure any services or products. The U.S. Army reserves the right to reject any or all responses. All information submitted in response to this RFI will be treated as proprietary.