OSW26BZ05-DV018: Artificial Intelligence / Machine Learning (AI/ML)-Based Radar Data Compression

Below is a brief summary. Please check the full solicitation before applying (link in resources section).

Quick Answer

OSW26BZ05-DV018 is a 2026 OSW-Reliance 21 SBIR topic under the DoW SBIR Program funding deep learning approaches, such as autoencoders or transformer models, that compress raw radar data into low-bit representations for efficient storage and transmission without losing radar utility. Phase I offers a base award of $314,363 over 12 months with a 20-page technical volume limit. A Direct to Phase II track is available at $2,095,748 over 24 months. This topic is ITAR/EAR restricted. Proposals are due September 23, 2026 and must be submitted through DSIP.

Overview

Next-generation radars, particularly synthetic aperture radar (SAR), generate enormous data rates that traditional image compression methods were never designed to handle. This topic is looking for neural compression approaches built specifically for the complex SAR data domain, using complex-valued neural network encoders and decoders that learn to compress raw pulses or range-Doppler maps into a compact latent representation with minimal distortion, while preserving the features that matter for detection and analysis.

Phase I is about designing and training autoencoder networks for radar data, benchmarking rate-distortion performance against classical compression methods on recorded datasets, and measuring the impact on target detection performance. Phase II moves the best-performing model onto embedded hardware for real-time compression, which likely involves quantization-aware training to reduce bit depth, and tests it against live radar feeds to evaluate image and track fidelity after decompression. Phase III integrates the technology into fielded radars or data links.

The dual-use angle here is broad. Any high-rate sensor stream benefits from this kind of compression, from unmanned aerial vehicle radar feeds to high-resolution weather radar data, which gives commercially minded proposers a real path to Phase III beyond DoW customers.

This is a strong fit for companies with deep learning expertise applied to signal processing, particularly teams with experience in autoencoders, transformer architectures, or complex-valued neural networks, combined with radar or SAR domain knowledge. Because this topic touches sensitive radar signal processing, it is ITAR/EAR restricted, and any proposed use of foreign nationals must be disclosed in detail.

Companies pursuing Direct to Phase II need documented proof-of-concept work that meets the Phase I objectives, independent of prior federally funded SBIR or STTR Phase I work, along with a clear commercialization discussion.

Funding and Timeline

Phase I base award: $314,363 Phase I period of performance: 12 months Phase I technical volume limit: 20 pages Direct to Phase II award: $2,095,748 Direct to Phase II period of performance: 24 months Direct to Phase II technical volume limit: 20 pages Proposal deadline: September 23, 2026 Submission portal: DSIP only, no other submission method is accepted Additional funding available: TABA funding, up to $6,500 for Phase I awardees and up to $50,000 per Phase II project Critical Technology Areas: Applied Artificial Intelligence and Quantum and Battlefield Information Dominance CMMC level requirement: Level 2 (Self) Export control status: ITAR/EAR restricted

Who Should Apply

Companies with strong applied machine learning teams, particularly those who have worked on compression, autoencoders, or generative models for scientific or sensor data, are a natural fit. Radar or SAR domain experience is a meaningful differentiator given the specificity of the technical objectives.

Frequently Asked Questions

What is the deadline for OSW26BZ05-DV018? The proposal deadline is September 23, 2026. Proposals must be submitted through DSIP before the topic closes.

How much funding is available for Phase I? Phase I offers a base award of up to $314,363 for a 12 month period of performance.

Can a company skip Phase I and apply directly for Phase II? Yes. This topic accepts Direct to Phase II proposals, but only from companies that can document completed proof-of-concept work meeting the Phase I objectives, performed independently of any prior SBIR or STTR federal funding.

What kind of AI approach is the government looking for? Deep learning based compression, specifically autoencoder or transformer architectures, ideally using complex-valued neural networks suited to SAR and radar data.

Is this topic export controlled? Yes. This topic is restricted under ITAR and EAR. Any proposed use of foreign nationals must be disclosed, including country of origin, visa or work permit type, and specific tasks assigned, and participation may be restricted.

What is the required CMMC level? CMMC Level 2 (Self) is the projected requirement for this topic.

What civilian applications does this technology have? Weather radar data networks and medical imaging streams are both cited as dual-use applications with commercial potential.

Is TABA funding available for this topic? Yes. Phase I awardees may request up to $6,500 in TABA funding, and Phase II awardees may request up to $50,000 per Phase II project.