ML Trajectory Deployment Optimizer

Compare deployment strategies using neural network trajectory estimation for optimal swarm deployment sequencing. Evaluates sequential, batch, greedy, and NN-guided approaches across Monte Carlo runs.

Deployment Parameters

Loading NN weights...
100 2,500 5,000
5 25 50
10 100 200 km/s
Strategies to Compare
10 100 200

Strategy Comparison

Compares 4 strategies deploying 500 units with 20 tugs over 50 Monte Carlo iterations.

Deployment Timeline

Run simulation to see deployment timeline

Propellant Usage by Strategy

Run simulation to see propellant comparison

Strategy Comparison

Run a simulation to compare deployment strategies

Simulation Methodology

This simulator models the deployment of collector units from an assembly node (L1 or L4) to evenly-distributed orbital slots around a heliocentric orbit. A trained neural network approximates delta-V costs for each transfer, replacing expensive Lambert solver iterations.

  • Sequential: Deploy units one-by-one in slot order, round-robin across tugs
  • Batch: Tugs carry 3 units per trip, deploying in angular clusters
  • Greedy: At each step, pick the nearest undeployed slot to assembly node
  • NN-Guided: Use NN delta-V estimates to globally optimize deployment sequence

Propellant consumption follows the Tsiolkovsky rocket equation with ion propulsion (Isp 3000s). Monte Carlo uncertainty comes from stochastic fuel efficiency (+/-5%).

Project Dyson — A volunteer-led nonprofit. All plans and research are publicly available.