DARPA Track at Big Distances with Track-Before-Detect (TBD2)
Executive Summary:
DARPA’s Strategic Technology Office (STO) is funding the Track at Big Distances with Track-Before-Detect (TBD2) program to develop advanced algorithms and payload designs for real-time space situational awareness in cislunar space. Selected teams may receive multi-million-dollar OTAs to build prototypes over a 15-month effort. Abstracts are due December 4, 2025, and companies should begin preparing materials now to meet the deadline.
How much funding would I receive?
DARPA anticipates multiple OTA prototype awards ranging from approximately $500K to $5M+, depending on technical scope, cost realism, and contribution to program goals. Larger awards are possible for high-complexity payload and algorithm development efforts.
What could I use the funding for?
1.Background
The goal of the TBD2 program is to enable continuous space-based detection and tracking of objects in cislunar space on relevant timelines. This effort will increase the safety of cislunar commercial and civilian traffic contributing to the peaceful use of space for the benefit of all nations and enabling a sustainable space ecosystem. To accomplish this, TBD2 seeks to advance the state of the art for signal processing algorithms so that when combined with commercial off-the-shelf (COTS) or quasi-COTS optical sensors and/or focal plane arrays (FPAs), they can a) detect and track faint objects at gigameter (Gm) distances, b) operate using available onboard processing capabilities, and c) do so on relevant timelines (within hours).
Figure 1: TBD2 seeks to extend space situational awareness beyond GEO to cislunar space
Existing space situational awareness (SSA) capabilities are primarily focused on objects in geosynchronous orbit (GEO) or closer. Extending SSA to cislunar space presents unique challenges as the distances are much greater and the volume of space needing to be scanned is ~1,200 times greater than GEO. Ground-based systems can combine large optics with complex, resource-intensive algorithms to enable cislunar detections, but are limited by their fixed location on the ground, inability to detect or track objects during hours of sunlight, and having to contend with weather and the Earth’s atmosphere. TBD2 seeks to solve this by moving the sensor to space, specifically to the Sun-Earth Lagrangian point 1 (SEL1) in order to negate any blinding of the sensor by the sun and enable a continuous view of most of cislunar space via a single sensor. To achieve this, TBD2 will require a novel approach to signal processing to detect faint objects (magnitude 23) at distances up to 2 Gigameter (Gm), while also minimizing processing time to the point that all cislunar space can be scanned within 12 hours. For a sensor at SEL1, sending image data to the ground for processing would require a continuous high-rate downlink that is impractical, if not infeasible, due to limitations in bandwidth, latency, energy budgets, and relative positions of the sensor and Earth. Therefore, TBD2 aims to develop or adapt signal processing algorithms that can be run in quasi-real time via onboard processing.
In addition to signal processing, TBD2 will also develop two distinct payload designs that optimize the combination of the signal processing with sensors/space-based compute platforms for two distinct scenarios: SEL1 and beyond GEO orbits. While SEL1 is of particular interest, alternate employment options could potentially enable a closer view of certain cislunar areas while also allowing for the detection and tracking of <1-meter objects that could endanger government, commercial or civil space operations. TBD2 will have three final deliverables:
1. The low complexity algorithm software implementation
2. Two payload designs that include the optics/sensor and compute platform combinations to be used for:
a. placement in SEL1.
b. placement in beyond GEO/cislunar orbits.
If successful, TBD2 will improve early warning capabilities for defense and civilian agencies who track potential threats and objects of interest originating from or transiting cislunar space, contributing to the safe and peaceful use of space for all nations. The fully developed signal processing algorithms capable of meeting program metrics and program goals and payload designs approved through Systems Requirements Review (SRR) constitute the Prototypes developed under the TBD2 Program.
TBD2 is a 15-month single-phase effort with two tasks.
Task 1 is to reduce the computational needs (and associated power consumption) of signal processing algorithms needed to detect faint moving objects at distances of millions of kilometers (km), and
Task 2 is to develop a payload design trade study that optimizes quasi-COTS sensors, onboard processors, and algorithms to achieve overall TBD2 goals.
1.2. Program Description/Scope
While many limitations of current approaches for cislunar SSA can be mitigated by placing sensors far from Earth (such as at SEL1), this introduces several technical challenges. Primarily, detecting and tracking faint moving resident space objects (RSOs) of 1 meter from SEL1 requires sensitivity levels capable of detecting objects as faint as 23 visual magnitude.
Achieving such sensitivity with quasi-COTS optics/sensors requires carefully optimized signal processing algorithms, which would traditionally be run via terrestrial compute platforms.
One theoretical path to achieving high sensitivity is through long integration times, stacking hundreds or thousands of image frames to boost signal-to-noise ratio (SNR), but this approach introduces latency which undermines achieving detection within appropriate timelines. While current synthetic tracking and track-before-detect algorithms can theoretically reach these sensitivities, they are computationally expensive—requiring around 300 Trillion Floating-Point Operations Per Second (TFLOPs) (FP32) to operate effectively from SEL1 with limited SNR loss. Addressing this compute need without compromising performance is the main goal of the TBD2 program.
There are several algorithms, generally belonging to the Track-Before-Detect family of Algorithms (TBDAs), developed for or adapted to the detection of faint moving objects in deep space – including near-Earth asteroids, main belt asteroids, and cislunar RSOs. While some methods linearly scale with the number of frames, pixels, and motion hypotheses, alternative strategies may yield sublinear or logarithmic scaling in some dimensions. TBD2 encourages the development of such efficient architectures to enable quasi–real-time onboard detection capability while also maintaining adequate performance.
TBD2 actively encourages exploration of innovative techniques, including but not limited to:
• Coarse-to-fine search methods (e.g., motion-aware pyramidal stacking).
• Radon transforms and their efficient approximations such as the Fast discrete X-ray Transform (FaXT).
• Probabilistic voting schemes to prune velocity hypothesis space over time.
• Techniques developed in the broader Infrared Search and Track (IRST) community combined with cislunar SSA algorithms to reduce complexity, e.g., exploiting their capability of treating hypothesized trajectories stochastically and pruning them early keeping complexity bounded, and their capability to operate at low SNR.
• Multi-sensor per platform designs, as multiple telescopes lower revisit time and reduce computational needs by reducing the number of pictures, shortening integration time, and reducing the number of hypothesized velocities.
DARPA is interested in the performance of TBD2 algorithms in several areas, and the government will provide data sets to test each of these areas individually as well as together. Some of the data sets provided will be real data from an optical sensor, others will be partially synthetic (i.e., real data with fake moving objects added to the data), and others will be totally synthetic data sets.
Data sets will be provided to each performer for “practice” with their approach, while additional data sets will be used for evaluating the algorithms. In general, the number of data sets provided for "practice" will not be sufficient for training Artificial Intelligence (AI)/Machine Learning (ML) algorithms, so AI/ML proposers are responsible for the training of their algorithms. If requested, the government team can provide guidance to each team on how to insert their own “fake” objects for training AI models. It is expected that, over the program period of performance, thousands of data cubes will be analyzed to collect performance statistics. The government has not finalized the data format yet, and the final data format may address discussions between selected performers and the government team.
At the midpoint and conclusion of the period of performance, performer algorithms will be evaluated on a government team-hosted platform to assess accuracy of detection and tracking of dim targets of various magnitudes and required computational power for quasi real-time execution. Each performer will need to provide an executable code that will be run on a common computer they will have access to. At the midpoint and final evaluations, approaches will be evaluated according to the program metrics (Section 1.5), but additional attributes will also be considered:
• Sensitivity vs. integration time tradeoffs.
• Peak TFLOPs and memory (gigabytes) required to store and process intermediate results.
• Astrometric accuracy (comparing output to limits imposed by optics, point spread function (PSF), and photon statistics; RSO velocity estimation accuracy, etc.).
• Weight and rough order of magnitude (ROM) cost of payload and bus.
In addition to the signal processing approaches, performers will be expected to develop two distinct payload designs: one for employment at SEL1, and another for employment in a beyond GEO/cislunar orbit. While SEL1 may be a good location to perform continuous cislunar SSA of ~1m sized RSOs, it is important to also explore additional options, such as orbits beyond GEO and around EMLs (Earth-Moon Lagrangians). These orbits offer closer views of the Moon and the Earth-Moon corridor thus allowing detection and tracking of smaller objects of magnitude 23 (10-20 cm at 200,000-400,000 km), as well as covering the remaining part of cislunar space that is obstructed from SEL1.
Four possible missions for the placements of a few TBD2 sensors include:
1. Monitoring the Earth-Moon corridor
2. Monitoring lunar orbits, including EML1 and EML2
3. Monitoring medium earth orbit (MEO)/GEO orbits
4. The small part of cislunar space that has an obstructed view from SEL1
The payload designs should suggest optimal optics and sensor, algorithm, and compute platform combinations for use at SEL1 and these additional orbits. Pursuing a single payload solution (optics and sensor plus compute platform combination) for the four cislunar missions identified above is strongly encouraged. These payload designs should consider aspects such as:
• Number of telescopes, optical aperture size, and sensor parameters
• SNR detection regimes (background-limited vs. read-noise-limited)
• Platform requirements for imaging at various integration times
• Required computational needs
• Consumed power
• Estimate of payload size, mass, power requirements
• Weight, power consumption, and ROM cost of payload and bus
For the payload design, factors such as payload mass and power consumption are of critical importance. For example, multi-sensor per platform designs are of interest as multiple telescopes reduce the number of pictures and shorten integration time, thus lowering revisit time and reducing the number of hypothesized velocities. Any proposed multi-telescope options would need to quantifiably justify the performance increase at the cost of mass and volume.
The payload designs will be evaluated near the end of the 15-month period of performance via a Systems Requirements Review (SRR) with the government team. This SRR will include examination and evaluation of the functional and performance requirements designed for the individual components (signal processing, computer platform, sensor) and overall payload of the two employment scenarios. The intent is that at the conclusion of the TBD2 program, the SRR-approved payload designs can be used to proceed with the initial system design by a transition partner.
Overall, at the completion of the 15-month period of performance for TBD2, the government’s expectation is to have prototypes of the fully developed signal processing algorithms capable of meeting program metrics and program goals and payload designs that have been approved through SRRs. This will pave the way for designing and building the payload after the end of the program.
Are there any additional benefits I would receive?
Beyond funding, TBD2 awardees gain significant strategic and reputational advantages:
DARPA Validation and Credibility:
Selection under DARPA’s STO signals exceptional technical capability and strategic relevance in defense and space innovation—often accelerating follow-on funding, partnerships, and investor confidence.
Enhanced Market Visibility:
DARPA-funded projects receive national-level attention in defense and aerospace circles, elevating recipients’ profiles as leading-edge space technology providers.
Ecosystem Access:
Participants collaborate with top experts in signal processing, optical sensing, and SSA, building direct connections to DoD transition partners and primes seeking flight-ready technologies.
Nondilutive Growth and Exit Value:
Because TBD2 is nondilutive federal funding, awardees retain IP ownership (with limited government-use rights), strengthening their valuation and commercial leverage for future acquisitions or private investment.
What is the timeline to apply and when would I receive funding?
Abstracts Due: December 4, 2025, 12:00 PM ET
Oral Presentation Invitations: by government request, estimated four weeks after abstract submission.
Awards Announced: Early 2026
Program Start: Upon award of OTA (15-month duration)
Funding is typically issued shortly after OTA negotiation and execution, following DARPA’s oral presentation evaluations and selections.
Where does this funding come from?
Funding is provided by the Defense Advanced Research Projects Agency (DARPA) under its Strategic Technology Office (STO) through Other Transaction Agreements (OTAs) authorized under 10 U.S.C. § 4022 for prototype projects.ds.
Who is eligible to apply?
Eligible applicants include:
Large and small businesses
Nontraditional defense contractors (per 10 U.S.C. § 3014)
Academic and research institutions (per 15 U.S.C. § 638(e)(8))
What companies and projects are likely to win?
DARPA will prioritize teams that demonstrate:
Proven expertise in signal processing, AI/ML, or SSA algorithms.
Ability to run real-time detection on space-qualified compute platforms with limited power (≤600W).
Designs that integrate quasi-COTS optics and sensors with innovative onboard processing.
Clear performance metrics and feasible payload trade studies for SEL1 and beyond-GEO orbits.
Collaborations between algorithm developers, optical engineers, and hardware integrators are strongly favored.
1.5. TBD2 Goals, Metrics, and Constraints
The objective of TBD2 is to enable continued space-based detection and tracking of objects in cislunar space within appropriate revisit timelines, thereby increasing the safety of cislunar commercial and civilian traffic and contributing to the peaceful use of space for all nations. The proposed concept is to place optical sensor(s) beyond GEO and use algorithms with reduced computational needs that run on available onboard processing to achieve this, and the program metrics are focused on key parameters for achieving space situational awareness. This includes metrics for detection range and sensitivity, revisit times, and onboard processing power consumption, combined with the ability to achieve positive detections while minimizing the chance of false detections.
In addition to the metrics, several constraints are provided in order to guide proposer solutions. The first constraint is that performers should assume a value of 20% albedo or less (i.e. assume no more than 20% of light is reflected from the RSO’s surface) for all potential RSOs. The second is to have performers assume that their processing time must be equal or less than the integration time when developing their payload trade studies. The third constraint limits any proposed optical aperture to a maximum diameter of 0.5 meters.
Are there any restrictions I should know about?
All work must be unclassified.
Cost-sharing may be required only for traditional defense contractors without nontraditional partners.
Export-controlled technologies must comply with U.S. export laws (ITAR/EAR).
How long will it take me to prepare an application?
For a first-time applicant, preparing a competitive abstract will likely take 40–60 hours in total.
How can BW&CO help?
Our team specializes in complex federal R&D proposals and can:
Triple your likelihood of success through proven strategy and insider-aligned proposal development
Reduce your time spent on the proposal by 50–80%, letting your team focus on technology and operations
Ensure you are targeting the best opportunity for your project and positioning your company for long-term growth.
How much would BW&CO Charge?
Our full service support is available for the Abstract for $4000. Assistance with Oral Presentation quoted upon invitation.
Fractional support is $300 per hour.
For startups, we offer a discounted rate of $250 per hour to make top-tier grant consulting more accessible while maintaining the same level of strategic guidance and proposal quality.
Additional Resources
View the Solicitation here.