AI-Driven Model Based Optimization to Enhance PWSA

The Space Development Agency (SDA) is developing the Proliferated Warfighter Space Architecture (PWSA), a network of small, low-cost satellites designed in functional layers: Transport, Battle Management, Tracking, Custody, Emerging Capabilities (Deterrence), Navigation, and Support. Many decisions must be made across all these layers to optimize mission performance globally. TensorX has demonstrated the power of AI-driven model-based optimization on the Tranche 1 Transport Layer (T1TL).  Even for a team that routinely improves global system performance by 20%, our preliminary results were staggeringly good.  By optimizing the cross-link topology alone, we demonstrated with a single cross-plane OISL equivalent performance to the baseline T1TL network with two cross-plane OISLs.  When we optimized the link topology globally, we achieved with three OISLs performance equivalent to the baseline topology with five OISLs.  If the cost of adding two OISLs, power, bus, control, etc. is $3M / satellite, that’s a network topology performance improvement that is equivalent to over $375M in additional T1TL hardware expense. The proposal will focus on optimization of the Tranche 2 Transport Layer (T2TL) including system design, R&D portfolio allocation, procurement, mission planning, and real-time adaptation. It will also include the combination of all Transport Layers. This will be accomplished by the power of a simulation-informed Mission Performance Model (MPM). A critical step in the optimization is configuring and tuning the Mission Performance Model (MPM) which can be envisioned as a weighted combination of the system-level Key Performance Indicators (KPIs) where the weights define the trade space. Products of the demonstration and analysis include the documentation of the MPM comprised of the relatively weighted KPIs including latency, congestion, mean time to access, and mean access time affecting simultaneous users. Design input parameters include but are not limited to orbital elements, network routing algorithms, and processing speed for the satellite network, signal acquisition, pointing, and transmission rates. Other inputs include transmission range and atmospheric height limits for the inter-satellite links.