OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Integrated Network Systems-of-Systems; Trusted AI and Autonomy OBJECTIVE: The objective of this topic is to develop secure, safe, reliable and economical approaches for the command and control of autonomous systems. This capability is important to the concept of Trusted Autonomy as described in the USD(R&E) Technology Vision for an Era of Competition . To ensure the security and reliability of the autonomous systems over a wide range of conditions a decentralized approach to command and control is needed since the alternative, a centralized approach to command and control, cannot be expected to perform reliably when communications are disrupted. The autonomous systems of interest in this topic are primarily low-cost surveillance, communications and delivery platforms operating in large numbers distributed over wide areas. A capability that enables command and control of large numbers of autonomous systems has military, public safety and commercial applications. The desired approach is an architecture that supports the use of large numbers of unmanned platforms that host sensors and communications links and can perform logistics support by delivering material. These unmanned platforms can be stationary or mobile ground vehicles, marine vessels, or aircraft. The platforms may be designed for long term unattended operations, especially for marine vessels. The ground and marine platforms may also be capable of automated launch and recovery of unmanned aircraft. The command and control system is expected to support a wide range of platform types and operating environments. It is also expected to operate with the existing command and control systems that manage the planning and execution of missions. The collection of autonomous systems must operate in accordance with legal and policy requirements. These requirements have been defined for US DoD systems and commercial systems have similar requirements. Command and control of these platforms requires an information infrastructure that supports strong identity management, secure messaging and workflows that include artificial intelligence and machine learning (AI/ML). The AI/ML workflows should use an information architecture that supports safety and reliability verification and testing through semantic descriptions of data flows and processing. DESCRIPTION: Industry trends point to increased use of autonomous systems in the future primarily due to economic benefits. These economic benefits favor large numbers of smaller, less expensive systems for many applications such as logistics, communications, and surveillance. There is a potential for a military organization to use large numbers of smaller, less expensive systems in support roles. Countering an adversary that adopts this approach could be difficult using current capabilities and may require developing a complementary approach. A method for effective command and control of large numbers of autonomous systems that is secure, reliable, resilient and economical will be needed. Command and control of autonomous systems is also relevant for commercial, public safety and scientific applications. A command and control system for large numbers of autonomous systems based on traditional technologies such as relational databases and centralized identify management could be more expensive and less reliable than a decentralized approach. A key component of economics will be the openness of the approach to allow for innovation and simplified integration. For the unmanned platforms to operate autonomously requires an information architecture that supports the integration of intelligence in the form of feature detection, course of action development and allocation of available resources. Decentralized databases, such as blockchain, could be used to create reliable and secure messaging and information storage. Recent developments in this area for decentralized finance (DeFi) have potential applicability, such as Layer 2 blockchains for improved performance and reduced cost, tokenization of data and identity to create Self-Sovereign Identification (SSI) and secure messaging. Integrating artificial intelligence and machine learning (AI/ML) into the command and control workflow is an important part of managing large numbers of autonomous systems. Commercial models for integrating AI/ML into command and control workflows include Uber's Michelangelo system . Michelangelo and other AI/ML workflows typically have a data ingest process, a feature detection process, recommendation process and a scheduler. Integrating AI/ML into a system that performs command and control of autonomous systems requires a higher level of verification and testing than a system like Uber's Michelangelo that can rely on the human operators to perform a validity check before taking action. PHASE I: Phase I proposals could include feasibility studies that examine architectures and technologies that support decentralized command and control of large numbers of autonomous platforms distributed over wide areas in various environments, including urban, rural and marine environments as well as proposals that focus on a specific aspect of this capability, such as cybersecurity, AI/ML workflows or verification and testing. Phase I proposals should describe what aspects of the problem the effort will be focused on and any previous work in this technology area. PHASE II: Phase II proposals could include evaluation of emerging technologies including performance and security assessments, through prototype development and demonstrations. Phase II proposals should describe what aspects of the problem the effort will be focused on and any previous work in this technology area. PHASE III DUAL USE APPLICATIONS: A potential Phase III application could involve the distribution of materials over a wide area with a set of collaborating autonomous systems, whether this is a commercial application delivering good to residences, or a military application delivering materiel to remote bases. No government furnished equipment or information or access to government facilities is expected to be required to complete these tasks. In lieu of demonstrations with large number of autonomous systems the proposers would likely perform simulations and possibly demonstrations with small numbers of simple autonomous systems. REFERENCES: 1. USD(R&E) Technology Vision for an Era of Competition 1-Feb-2022, pg. 4; 2. https://www.airuniversity.af.edu/JIPA/Display/Article/3091254/taming-the-killer-robot-toward-a-set-of-ethical-principles-for-military-artific/; 3. https://media.defense.gov/2019/Jun/18/2002146749/-1/-1/0/JP_001_COOK_TAMING_KILLER_ROBOTS.PDF; 4. https://media.defense.gov/2021/May/27/2002730593/-1/-1/0/IMPLEMENTING-RESPONSIBLE-ARTIFICIAL-INTELLIGENCE-IN-THE-DEPARTMENT-OF-DEFENSE.PDF; 5. https://www.ibm.com/blogs/blockchain/category/trusted-identity/self-sovereign-identity/; 6. https://www.uber.com/blog/michelangelo-machine-learning-platform/ KEYWORDS: command and control; decentralized; autonomous; blockchain
