Mr. Sriram Jayasimha, Founder, Signion, and Adjunct Faculty, SPCRC, IIITH gave a talk on AI in LEO Constellation-Based Mobile Communication Network on 19 February. 

Managing a Low Earth Orbit (LEO) Satellite Constellation for mobile communications presents complex challenges in decision-making and resource allocation. These challenges necessitate deploying autonomous AI-driven agents capable of optimizing network performance in real time. The agents leverage real-time telemetry data to predict network states, detect anomalies, dynamically allocate resources, and ensure seamless service continuity while maximizing operational efficiency and revenue.

• Resource Allocation: A key objective is the dynamic allocation of multi-beam resources to ground cells, with power generated by the satellites and considering constraints such as power amplifier saturation, feeder link bandwidth availability, service-level agreements (SLAs), and cross-border regulatory policies governing spectrum usage. Additionally, the system must optimize capacity while balancing Quality of Service (QoS) and revenue maximization. Such a problem is well-suited for reinforcement learning (e.g, evolutionary algorithms, genetic algorithms, and deep Q-learning), with scalable multi-agent execution frameworks.

• Constellation Management: Constellation management entails adaptive control of phased array (PA) payloads based on multiple dynamic factors, including payload health, satellite battery state-of-charge, mean sun angle exposure, and inter-satellite coordination to prevent service disruptions. Additionally, intelligent beam handoff decisions are required to transition service areas between satellites while mitigating power constraints and optimizing cell elevation angles. These decision-making problems are effectively addressed using ensemble learning techniques, such as random forests of decision trees, logistic regression, or Bayesian optimization.

Both AI-driven agents operate autonomously yet must coordinate to achieve business and operational goals, ensuring efficient spectrum utilization and regulatory compliance. A crucial capability is agent-to-agent explainability, enabling transparency in decision-making to prevent conflicting actions.

Sriram Jayasimha founded Hyderabad-based Signion in 1987 with a 37 year track record of IP creation and product design for the global telecommunications markets. From May 1981 to June 1987, Sriram was a member of the technical staff at GTE Government Systems Corp. in Needham, MA. He has been a consultant to Compusonics Corporation, Bharat Electronics Limited (Central Research Laboratory), and Aware Inc., Cambridge, MA.   Since retiring from Signion in 2019, he is currently the Chief Scientist of AST Space-Mobile, which plans to bring global connectivity to mobile phones using a constellation of large aperture satellites. He has authored 42 peer-reviewed journal and conference publications in an array of digital-signal-processing applications, including mobility, satellite communications, modem, VLSI, and instrumentation signal processing. A named inventor on 28 unique U.S and other European, Japanese, Korean, and Australian patents, Sriram was a fellow of the Center of Advanced Engineering Study at Massachusetts Institute of Technology. He earned his B.Tech and M.S degrees in electrical engineering from the Indian Institute of Technology, Madras, and Rensselaer Polytechnic Institute, respectively. He is currently an adjunct/ honorary professor at IIIT Hyderabad.

February 2025