Yasir Latif, Space Protocol; Samya Bagchi, Space Protocol; Latha Madhuri Pratti, Space Protocol; Harvey Reed, MITRE; Ruth Stilwell, Aerospace Policy Solutions, LLC; Greg Furlich, University of Colorado Boulder
Keywords: SDA, Reinforcement Learning, Agent AI, Battle Management Software, SDA TAP Lab
Abstract:
The Imperative for Adaptive Space Battle Management
The proliferation of space debris, adversarial threats, and the dynamic nature of orbital operations demand Battle Management Systems (BMS) that balance rapid decision-making with rigorous traceability. Current systems often operate as black boxes, lacking mechanisms to audit decisions, refine performance over time, or explain actions to human operators – critical gaps in high-stakes scenarios where errors cascade catastrophically. This work addresses these limitations by integrating reinforcement learning (RL), blockchain-based traceability, and causal analysis into a unified framework designed for Space Domain Awareness (SDA).
A Synergistic Architecture for Trust, Adaptability, and Accountability
We propose an architecture that combines:
Blockchain-Immutable Traceability: Every data transformation across the SDA pipeline—from sensor ingestion to response recommendation—is logged to a distributed ledger with cryptographic hashing. This creates an audit trail resilient to tampering, essential for post-event analysis and compliance with international space treaties.
Reinforcement Learning Oversight: A meta-layer of RL agents monitors subsystem interactions (target modeling, hostility assessment, command prioritization) using multi-objective reward functions. These agents optimize for mission success metrics while penalizing decisions with low traceability compliance or unexplained confidence variances.
Verification Learning: Independent validators cross-check subsystem outputs against curated test datasets (e.g., historical collision events, simulated adversarial maneuvers) to detect model drift or adversarial data poisoning.
OODA Loop UI: Insights from traceability, oversight AI and verification learning are integrated at various touch points in an Observe, Orient, Decide, and Act (OODA) loop to enable the operator to make better decisions under time pressure
The system uniquely integrates causal analysis to map decision pathways, assigning stepwise confidence scores that inform both machine learning updates and operator briefings. A chat interface allows operators to query the rationale behind high-risk decisions (e.g.,“Why was Sensor X prioritized over Y during Event Z?”), with responses grounded in blockchain-verified logs and RL-derived correlation models and informing the operator in an easy to interpret and execute OODA loop framework.
Implementation and Validation in the SDA TAP Lab
The SDA Tools, Applications, and Processing (TAP) Lab was formed in Colorado Springs, Colorado in 2023 bringing together industry, academia, and government to deliver a Space Battle Management System. Space Protocol has adopted NIST IR 8536 “Supply Chain Traceability: Manufacturing Meta-Framework” to provide traceability for the TAP Lab’s BMS. Data-to-decision traceability allows comprehensive and auditable documentation of the lifecycle of data, encompassing its origin, transformation, movement, and utilization, as well as the rationale and influencing factors behind decisions using that data. The TAP LAB BMS is organized along seven discrete event workflows that include launch, reentry, proximity, maneuver, link change, attitude change, and separation. Each workflow represents an event that is of interest to the operator and a critical decision point in the system.
Space Protocol has adopted the traceability framework from the original supply chain context to data-to-decision traceability within the context of the SDA TAP Lab in general for each workflow in particular. Kafka, the common data back-bone, enables a unified data exchange mechanism, providing information about each data point and its transformation, providing a link in the traceability records. These records are stored using a quantum resistant distributed ledger. The data gathered via traceability enables:
Graph-Based Auditability: Novel algorithms parse traceability records and reconstructs decision chains with millisecond latency, even for 10^6-node graphs, enabling rapid forensic analysis and real-time introspection.
RL-Driven Optimization: Soft Actor-Critic (SAC) based agents utilize historic data and are applied to each event workflow in the BMS, such as false track associations, satellite identification, and friend vs foe assessment.
Human-AI Collaboration: AI-Agent and chatbots provide detailed system analytics in a human focused and interactive manner. Custom models are employed on premise for safety and security reasons.
Operator-awareness: The use of OODA loop based visual interface into the system ensures that operator familiarity and interactivity with the system.
This work bridges critical gaps between autonomous decision-making and human accountability in SDA systems, providing a template for responsible AI integration in orbital operations. The TAP Lab implementation demonstrates practical implementation of the system from the ground up and will mature as the BMS being developed matures.
Date of Conference: September 16-19, 2025
Track: Space Domain Awareness