Day 2: Space Technology
Cost to go to orbit is decreasing. The number of satellites will increase exponentially very soon. Inspired by YC’s RFS, I spent 8 hours exploring how we can support the future of space communication.
The problem
The number of satellites will grow exponentially as companies like SpaceX lower the cost of launching and operating constellations. This means the demand for low-latency, high-bandwidth communication from space to Earth and vice-versa will only increase.
Currently, many satellites wait to pass over ground stations before they can transmit data back to Earth. This creates significant delays:
Most low Earth orbit (LEO) satellites complete a full orbit in ~90 minutes, with only a few minutes of connectivity during each pass over a ground station.
As the number of satellites grows, the existing approach will face bottlenecks.
We’re approaching a future where space communication systems need to function like ISPs on Earth — seamless and scalable.
This will allow:
Better command of satellites to respond to changing conditions
Magnitudes more data to be sent from satellites to the ground and back
Approach
What I set out to do
Simulate a satellite network with realistic orbits and interconnections.
Build a routing system to route data between satellites and ground stations.
Figure out the technical viability and hardware specs.
Technology used
Graph algorithms
Physics simulation for orbital motion
Python, Plotly, Matplotlib
How I approached the problem
I simulated a Walker constellation of satellites in 3D space, each orbiting Earth in low earth orbit (LEO, ~500km altitude).
Built a connectivity graph between satellites based on line-of-sight, range and number of peers.
Create a routing system to let data hop the satellites to get to the closest ground station.
Insights, Build Process & Technical Details
Method 1, building our own constellation*:
How do we get data from any satellite down to the ground in milliseconds? Taking inspiration from our networks, I built a simulated constellation of satellites where data should be able to hop between them and finally down to a satellite that is positioned over a ground station.
To get coverage of the Earth, I used the Walker constellation geometry with 33 planes of 11 satellites each. I chose an inclination of 53° to balance polar and equatorial coverage. Let’s call this constellation Skynetwork (SN).
We should use gimbaled lasers/receivers, similar to Starlink, for high-bandwidth communication between satellites in this network. These are extremely high bandwidths, setting a record of 13,160 Gbps compared to a millimeter wave which was only 40 Gbps. Google X is working on bringing this technology to ground systems.
In my simulation:
Users’ satellites send data using a dedicated laser transmitter, beaming data to the nearest SN satellite.
The data is routed between satellites in our constellation, dynamically balancing the load, until it reaches a satellite positioned above a ground station (using graph traversal algorithms and bottleneck calculations).
Currently using bandwidth as a constrain, can be expanded to factor cost/power
Once on earth, we use traditional networks to get data to the end user.
In Fig 1, on the left, we have a 3D view of our satellite constellation and 20 different satellites sending data to the ground on it (different colors). On the right, we show the 2D projection of these paths using Lat/Long.
The physical travel time of a laser-based system to circle half the Earth would be ~75 ms and most paths are shorter than that so latency is greatly improved.
Method 2, peer-to-peer:
Since building an entire constellation like Starlink is capital-intensive and slower to scale, I explored an alternative inspired by decentralized peer-to-peer networks.
Instead of operating the dedicated SN constellation, this model leverages the thousands of satellites being launched by startups, the private sector, and governments. Satellites equipped with our networking hardware become part of the decentralized system.
As more satellites join the network, its coverage and capacity grow organically, improving the service for everyone.
The idea is to let satellite operators add the gimbal/networking hardware to their satellites to let data pass through as it hops from location to location (as simulated above).
Participation Incentives:
Satellite owners can send/receive their own data over the network.
Owners earn revenue from the bandwidth their satellites pipe.
Technical considerations:
A dedicated constellation will suffer from the UDP problem — satellites sending data to our network won’t have instant feedback on whether the data was successfully transmitted. This is because to have feedback, our SN satellites would have to also transmit data back to them, reducing how many users each satellite can serve at once. The decentralized network can help solve this issue.
To reduce momentary live connection losses when changing which satellite a user is targeting (as they move out of range), we will have a redundant system. The user will send data at the same time to the closest SN satellite and also the one they will switch to when the first is out of range. Even then, there will have to be some feedback system at the end stations to manage the loss. This is also helped by the decentralized system since there can be instant feedback on if data was received.
Future Directions
Real-Time Optimization: Integrate machine learning to predict satellite congestion and optimize data flows.
Compute: could serve as a data pipeline for in-orbit compute clusters, enabling training and inference workloads directly in space.
Coverage server satellites: even with decentralized we can send up some of our own satellites to ensure coverage.
Key Takeaways
Satellites are becoming part of a global internet backbone: As satellite constellations grow, the demand for seamless communication between space and Earth will rise.
Two approaches to solve the problem: a dedicated constellation or a decentralized peer-to-peer model
Satellites are dynamic nodes in a global network: Unlike current ISPs, satellite networks must adapt to constant motion and changing connectivity
Decentralization for networks: A peer-to-peer network avoids the high costs of owning a constellation while providing a decentralized user-owned solution
Final Thoughts
While designing Skynetwork, I discovered that companies like Kepler Communications are already working on similar dedicated constellation solutions. However, there is significant opportunity in the peer-to-peer model, especially as satellite launches accelerate in the coming years.
Check out updates on X ➡️ #AdventOfYC | @AshrayGup
Lovely visuals. I'm curious, what's the benefit of having compute on the satellites themselves if the latency is low?