@spacesustainableai
Passionate about AI in Space, Military SpaceTech and Sustainable AI. I read a lot and post a blend of literature excerpts and my thoughts PhD from Nanyang Technological University, Singapore 🇺🇦Ukrainian in Singapore🇸🇬
Pro's and con's of Sustained Space Maneuver and difference between strategic (dubbed "reactive") and tactical maneuvers to ensure supporting USSF Competitive Endurance in his paper
csps.aerospace.org/sites/defaul...
#MilitarySpacetech #Space #SpaceAI #Spacewarfare
We published a new commentary on state of SpaceAI in Singapore and Southeast Asia! We spoke about sustainability on Earth and in orbits, use of AI for tropical climate dynamics and maritime security, and the power of cross-country collaborations in space.
rsis.edu.sg/rsis-publica...
3/3 China's Approach to Deterrence
💡Emphasizes soft‑kill tools in early stages, to degrade without escalation
💡Maintains redundant constellations to signal resilienc, raising the bar for strikes
💡Reserves kinetic ASATs for last-resort, high-intensity scenarios
2/3 Rivalry with the US
🛰️Chinese analysts identify US reliance as an asymmetric leverage point. US relies heavily on GPS, ISR, and comms satellites, becoming a primary target for attacl. Beijing also faces vulnerability with ~650 satellites of its own
1/3 China is building a sophisticated counterspace toolkit:
📡Soft-kill weapons: lasers, jamming, electronic/cyberattacks, reversible and deniable
☄️Hard-kill ASATs and resilient satellite constellations ensure redundancy
www.tandfonline.com/doi/epdf/10....
The proliferation of missile threats has made air & missile defense interceptors the table stakes for entry into future conflicts, forcing the DOD to either buy the necessary interceptors or fold on its regional interests.
More from CSIS's Missile Defense Project: www.csis.org/analysis/dep...
Recently joined the panel discussion "Conscious consumption of AI" by AlterCOP, SG, and spoke about:
🍀 AI as catalyst and inhibitor of sustainability
♻️ GHG emissions + resource footprint of AI models (Mistral report)
🪛 Optimizing models, hardware, architecture & data centers
📃 End-user hacks
3/3
🛰️ Why is DevOps rare in space?
✅ Onboard software = static & hard to update
✅ Missions demand extreme reliability & certification
But imagine this:
🚀 CI/CD + MLOps in orbit
✔ In-orbit software updates
✔ Continuous AI training & monitoring
✔ Lower costs, higher agility
2/3
AI for Autonomous and Adaptive Systems.
1. Onboard decision-making - adjusting orbit and power.
2. In-flight orbit determination - using optical sensors and Kalman filters for CubeSats.
3. Environmental adaptation - sensors + ML for radiation, thermal extremes, and collision avoidance.
1/3
Continuous Integration and Deployment can revolutionize satellite operations by enabling in-orbit updates, faster deployments, and improved reliability.
Why are we stuck with static onboard software? MLOps in space for continuous training & monitoring is the solution
doi.org/10.1145/3757...
On the American side, corrupt oligarchs and traitors to the United States.
On the Ukrainian side, leaders doing their best to save their country from the russian terrorist state and the Trump regime.
(Russian agent Steve Witkoff has pulled his chair farthest from the table.)
🔎New satellite imagery from Severodvinsk shows the Russian nuclear-powered submarine Khabarovsk in full size for the first time.
🔗Read details: militarnyi.com/en/news/khab...
4/4
The Ministry of State Security runs China’s most covert cyber ops. It hires contract hackers via front companies. Provincial MSS branches run ops with autonomy, making attribution nearly impossible. It’s a decentralized, deniable, and deeply embedded system.
3/4
Data is China’s strategic ammunition.
80% of Americans have had their data stolen. Not just for spying, but to train AI, identify vulnerabilities, run influence ops.
The 2014 Anthem breach used malware from a student hacking competition. The weaponization of young talent is chilling.
2/4
Volt Typhoon is China’s silent saboteur. This APT group embedded itself in U.S. critical infrastructure - water, power, transport - not for spying, but for disruption.
It’s pre-positioning for a Taiwan contingency. A 24-hour delay in U.S. response could tip the balance.
#cyberwarfare
1/4
Understanding China’s cyber warfare strategy, based on #CSIS report www.csis.org/analysis/pla...
China doesn’t separate war from peace in cyberspace.
The PLA’s doctrine treats cyber operations as part of espionage, disruption, and influence. It’s not hacking - it’s shaping perception and control
7/7
Energy-aware AI models predict solar power availability and battery health, adjusting payload operations to maximize mission life without human intervention. AI-enhanced thermal control predicts heat loads and dynamically adjusts radiators and heaters, optimizing energy use.
6/7
Onboard scheduling uses AI-driven optimization to allocate observation tasks dynamically. Deep Q-Networks adapt to changing priorities and orbital constraints in real time. AI-based fault recovery employs hybrid reasoning: rule-based systems for known issues and ML model for unknown anomalies.
5/7
AI-powered Event-Triggered Control reduces communication load. Instead of constant telemetry, satellites transmit only when anomalies or threshold events occur. Anomaly detection in spacecraft telemetry uses unsupervised learning to detect sensor drift or component failures before breakdowns.
4/7
Federated Learning allows satellites to train AI models collaboratively without sharing raw data, only model updates (like gradients or weights). This preserves bandwidth and security while improving onboard autonomy for heterogeneous constellations.
3/7
Graph Neural Networks (GNNs) are applied to satellite constellations for distributed decision-making. Each node (satellite) shares local state, enabling coordinated collision avoidance and resource allocation. GNNs can surface which nodes/edges drove an alert, aiding root cause analysis
2/7
Onboard AI uses reinforcement learning (RL) for autonomous navigation. RL agents optimize thrust and attitude control under uncertain dynamics, reducing reliance on ground commands. It also handles actuator failures, maintaining stability even when traditional control laws fail.
1/7
Types of ML techniques and neural networks widely used for AI in Space
arxiv.org/pdf/2406.14297
#SpaceAI
Right to the point. Observing so many russians in Singapore working for critical infrastructure, cybersecurity companies, shipping and logistics, openly spreading their imperialistic mindset, makes me wonder where the national guardrails are
During the Russian missile and drone strike on October 5th, at least three Chinese reconnaissance SAR satellites flew over western Ukraine, the main target of the attack.
🔗Read more: militarnyi.com/en/articles/...
4/4
Space Force plans to integrate cyber detection with orbital maneuvering and network segmentation. If a satellite is compromised, it can isolate itself, reroute traffic, and even reposition to avoid further risk, demonstrating a layered defense approach.
3/4
AI is key to defending satellites. Machine learning models analyze telemetry for patterns that indicate cyber tampering, like unexpected power draws or command sequences, helping operators respond before damage spreads.
2/4
Traditional cybersecurity tools don’t work well in orbit. US Space Force is investing in lightweight, radiation-hardened monitoring systems that can run autonomously, flag suspicious commands, and isolate compromised components in real time.
1/4
Satellites are prime cyber targets, and attacks often go unnoticed until systems fail. Space Force is deploying onboard sensors and AI-driven anomaly detection to spot intrusions early before hackers can disrupt comms or navigation.
#MilitarySpaceTech
#SpaceAI
