The optimization problem: maximize the number of user-hours connected given constraints on battery (solar recharge rate), wind prediction error, and balloon longevity. This became a partially observable Markov decision process (POMDP) with >10^6 state variables.
Loon’s envelope used helium. To lift a 15 kg payload (electronics + batteries) plus a 15 kg envelope, the balloon required displacing ~30 kg of air. At 20 km altitude (pressure ≈ 50 hPa), the volume needed is:
[ V = \frac{m_{air}}{\rho_{strat}} \approx \frac{30 \text{ kg}}{0.088 \text{ kg/m}^3} \approx 340 \text{ m}^3 ]
Loon required —a fully sealed, rigid envelope that maintains internal pressure higher than the external atmosphere at all times. The challenge: as the sun heats the balloon, internal pressure rises, stressing the polyethylene film.
The optimization problem: maximize the number of user-hours connected given constraints on battery (solar recharge rate), wind prediction error, and balloon longevity. This became a partially observable Markov decision process (POMDP) with >10^6 state variables.
Loon’s envelope used helium. To lift a 15 kg payload (electronics + batteries) plus a 15 kg envelope, the balloon required displacing ~30 kg of air. At 20 km altitude (pressure ≈ 50 hPa), the volume needed is:
[ V = \frac{m_{air}}{\rho_{strat}} \approx \frac{30 \text{ kg}}{0.088 \text{ kg/m}^3} \approx 340 \text{ m}^3 ]
Loon required —a fully sealed, rigid envelope that maintains internal pressure higher than the external atmosphere at all times. The challenge: as the sun heats the balloon, internal pressure rises, stressing the polyethylene film.
