TECHNOLOGY AREAS: Air Platform; Battlespace The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), 22 CFR Parts 120-130, which controls the export and import of defense-related material and services, including export of sensitive technical data, or the Export Administration Regulation (EAR), 15 CFR Parts 730-774, which controls dual use items. Offerors must disclose any proposed use of foreign nationals (FNs), their country(ies) of origin, the type of visa or work permit possessed, and the statement of work (SOW) tasks intended for accomplishment by the FN(s) in accordance with section 3.5 of the Announcement. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. OBJECTIVE: The Air force Research Lab Munitions Directorate (AFRL/RW) aims to create a framework to push real time updates to weapon systems. Currently, weapon behaviors are not informed based on real-time data. The objective of this effort is to demonstrate the performance to be gained by using data gathered in real time to enhance weapon performance. Data from the field should continuously influence performance models, and be leveraged by artificial intelligence based analytics to increase the capability of weapon effectiveness in adversarial environments. Swappable AI-based components will improve autonomous decision making by impacting key areas such as mission planning, mission data, and collaborative strategies for targeting, guidance, and estimation. The desired solutions should leverage the Weapons Open Systems Architecture (WOSA), increasing the opportunity for innovative solutions from a variety of vendors that may have niche expertise in specific WOSA domains. By using a modular architecture, Software Enable Weapons solutions enable tailored effects and performance that will adapt to real-world observed capabilities. The overall goal is to develop novel collaborative machine learning-based control and decision-making strategies for multi-agent systems that can dynamically adjust and collaborate to achieve system-level missions in adversarial and uncertain environments. DESCRIPTION: Currently, weapons are optimized to have a high-level of performance in very specific use-cases or a low-level of performance across a wide variety of use-cases. The refresh cycle for weapons is slow and weapons are not able to be tailored dependent on the situation, nor are their capabilities adaptable to a changing operational environment. The overall vision of Software Enabled Weapons is to create an app-store like environment where WOSA compliant, modular software apps can be modified to enhance the weapon's effectiveness. This SBIR effort looks to find data-driven, modular solutions that will ultimately be integrated within a weapon's app-store. Algorithms delivered as WOSA compliant applications could be created for any domain within the WOSA framework. This could include changes in the flight control system, the fuze behavior, guidance and navigation algorithms, seeker performance, health monitoring, networked collaborative autonomy behaviors, and others. Some of the challenges to consider are algorithmic scalability assuming a constrained communications environment, varying and limited compute and processing power, decenetralized learning, uncertainty, multi-modal adversary, contested environments, physics-informed learning, complex relationships between platforms. However, the scope of this effort is not to solve the challenges of the software update pipeline, but instead to focused on developing the potential apps that would be pushed to the weapon. PHASE I: As a D2P2 effort, a proposer should already have integrated machine learning algorithms on flight, or flight equivalent hardware, and have a demonstrated ability to execute a Continuous Integration continuous Development (CICD) pipeline to update their algorithms. The performer should have experience using generative artificial intelligence or reinforcement learning techniques. They should have demonstrated executing reinforcement learning based decision-making strategies in real-world scenarios. PHASE II: The expectation for this effort is for the performer to demonstrate data-driven weapon behavior solutions that are dependent on observed data in the environment. The performer must be able to show adaptability to unexpected situations in the adversarial environment using sparse data sets. The Phase II effort should demonstrate using artificial intelligence techniques to update performance models of weapons or to create novel approaches to weapon behaviors. The effort should also explore using generative artificial intelligence or reinforcement learning to enhance heterogeneous weapon selection, sensor placement, and guidance strategies to optimize towards mission objectives. The objectives for this effort is to deliver applications that can be integrated into the WOSA framework, including demonstrations of how the behavior of a weapon would adapt as new data is observed. The performer is encouraged to use open-source data sets for developing algorithms that may be applicable to the weapons domain. The performer should look to leverage representative simulation environments, such as AFSIM, or other frameworks to train agents in representative environments. The applications built for different WOSA domains (such as navigation, communications, guidance, etc...) should be demonstrated in a WOSA compliant environment. The application will be able to take in new data and adjust its algorithm based on new data received from the field. The application will demonstrate an increase in effectiveness after ingesting new data. The application will possess the capability to be tailored to a scenario as new information is revealed to it, or have specific parameters for the scenarios for which it is optimal. The applications provided should be hardware agnostics, weapon-centric, modular, and reusable to be accepted for integration during Phase III. PHASE III DUAL USE APPLICATIONS: Phase III of this effort would include application integration into the Software Enabled Weapons pipeline, to be flown on a modular weapons system surrogate. The expected TRL at the end of Phase III will be TRL 7 after a end-to-end prototyping demonstration. The applications developed for the Software Enabled Weapons program would be expected to go through proper software certification channels to ensure application compliance with DoD regulations. REFERENCES: 1. Voyager: An Open-Ended Embodied Agent with Large Language Models https://arxiv.org/pdf/2305.16291 2. Self-Supervised Policy Adaptation During Deployment https://bair.berkeley.edu/blog/2021/02/25/ss-adaptation/ 3. NVIDIA Isaac Sim https://developer.nvidia.com/isaac/sim KEYWORDS: Artificial Intelligence; autonomous vehicle; network collaborative autonomy; open system architecture; modular weapon system; agile development; mission planning.
