Description
Objective: Develop and integrate a distributed Artificial Intelligence (AI) technology that can collaboratively command and control a multi-agent Group 1 or Group 2 Unmanned Aerial System (UAS) swarm to defend an area against a numerically superior enemy swarm. AI system must be able to interface with existing, standard platform autonomy and perception systems. System must also not rely on a centralized control node. Description: Develop and integrate distributed Artificial Intelligence (AI) technology that can collaboratively control a multi-agent Group 1 or Group 2 Unmanned Aerial System (UAS) swarm to defend an area against a numerically superior attacking enemy swarm. The vast majority of counter-UAS systems are optimized for a 1 vs 1 scenario, in which the interceptor UAS seeks to destroy, degrade, disable, or capture a single enemy UAS. This approach typically relies on a sensor package and kinetic or non-kinetic effector optimized to degrade/destroy a single enemy UAS of a specific class (size, range, speed, etc.). To defend an area against a numerically superior enemy swarm, individual UAS platforms must collaborate to determine the optimal strategy for many individual 1 vs N scenarios. Individual UAS platforms must demonstrate the ability to target a cluster of enemy platforms through AI algorithms and active or passive inter-drone communication for targeting information from other friendly platform perspectives. The UAS platform employed can be an off-the-shelf OEM or custom-built. Key system attributes include: Must be able to collaborate across a homogeneous set of Group 1 or Group 2 UAS platforms to actively inform each friendly platform (given permissive network environment) of enemy UAS location, velocity, track, etc. Must be able to execute algorithms under extreme SWaP-C constraints with a total compute payload under 2 lbs. Must be able to passively collaborate and achieve similar, but degraded performance within a non-permissive network/communications environment. A single UAS platform must demonstrate the ability to degrade/destroy N enemy UAS within a range of 10 meters through either kinetic or non-kinetic effectors. Friendly UAS swarm must be able to severely degrade the combat power of the enemy to a fraction of X% of its original size. Although 100% degradation of the enemy swarm is ideal, depending on the degree of asymmetry, it may not be realistic. Therefore, a target final enemy combat power goal is achieved from the degree of enemy/friendly asymmetry. As shown in Figure 1, in a scenario where there is no initial asymmetry and our 1 v 1 capabilities are superior, the enemy should retain 0% of its original combat power. However, if that enemy/friendly initial combat power ratio were 4/1, final enemy combat power might be 75% of its original. The ideal curve is one in which final enemy combat power is 0% regardless of the initial asymmetric combat power advantages the enemy possesses. This effort is not designed to create, design, or deliver a new UAS platform as the end item. Rather, it is meant to develop technology that will leverage the existing capabilities of OEM drone platforms or, if necessary, custom-built drones by the performer. The key deliverable is a suite of AI and other software algorithms that continuously plan and take optimal actions in a decentralized manner. The algorithms run on each individual UAS platform, take advantage of active communications with other friendly platforms when operating in a permissive network environment, but can still operate under a denied or degraded network environment by communicating passively. Keywords: asymmetric; collaborative; denied; artificial intelligence; distributed CMMC Level: Level 1