High Altitude Platform Stations (HAPS) are effective tools for communication and surveillance because they operate from the stratosphere much closer than a satellite, typically about 20 km (12 miles) above Earth. There are two types of HAPS: light-than-air (LTA) HAPS, such as high-altitude balloons, and heavy-than-air (HTA) HAPS, such as fixed-wing aircraft.
LTA-HAPS are generally solar-powered unballoons that use gases (lighter than air) such as hydrogen or helium to maintain buoyancy. But they are capricious of the wind. The way LTA-HAPS maintains its position depends on complex interactions between naturally occurring phenomena and AI.
HAPS combines the flexibility of an aircraft with the durability of a satellite. As such, it can be used for telephone and Internet services, overhead communications such as radio broadcasting, earth observation, military intelligence gathering, reconnaissance, and surveillance (IRS).
Loon, a subsidiary of Alphabet Inc., has spent nine years with Google Brain developing the LTA-HAPS technology that makes this possible, to bring internet connectivity from the stratosphere to hard-to-reach places around the world.
There are obvious challenges in relying on solar-powered, propellerless, unmanned balloons to provide these types of services. The main one is the weather, which cannot be controlled. Balloons must be able to withstand high winds and large temperature changes over time while providing a constant connection to the ground below. Weather forecasts are often wrong, so relying solely on them is dangerous.
Additionally, balloons do not have an infinite on-board energy source. Relying on the sun to power both navigation and communications, automatically correcting a balloon if it is blown off course is a waste of precious power.
Research scientist Marlos Machado was part of the team that developed the technology that allowed the balloon to perform as designed and with maximum efficiency.
“We want the balloon to be in a specific position,” Machado said. “But there’s a catch. The problem with these balloons is that they don’t have propellers. They’re the only way to [can] Sailing in the stratosphere is riding the wind. “
“Managing the wind” refers to how the balloon uses wind strength, altitude, and direction to maintain a desired position (called stationkeeping) or navigate to a new location.
These balloons are huge, the size of a tennis court. In the stratosphere, winds are blowing in different directions at different altitudes and speeds, and the balloon is confined to her two directions, up and down, in that it controls the movement of the balloon itself.
“The idea is that there is a lighter-than-air gas in the balloon. [will] Get up,” said Machado. “If you want the balloon to sink, you just pump air into it.
If you introduce a lighter-than-air gas into the bladder inside the balloon, the balloon will rise in the wind flow and must move to the correct position. If the balloon needs to be lowered so that it is carried in the other direction by the wind, either pump ambient air into a fixed envelope inside the base of the balloon to provide ballast, or open the valve on the LTA bladder. Exhaust some of the gas so that the balloon becomes heavier.
To gain some control over this somewhat limited maneuverability, researchers developed a reinforcement learning AI that rewards balloon behavior. The simulator combined weather forecasts with observed weather data and Gaussian processes to provide the best wind predictions. If the balloon responded “correctly”, it was rewarded. However, sometimes the wind wins.
“Sometimes you can’t do anything,” Machado said. “When the wind direction is bad, we have no choice but to wait until a good wind blows.”
Ultimately, the research was successful. In 2017, the team piloted balloons to provide emergency connectivity to hundreds of thousands of people after flooding in Peru and a massive hurricane in Puerto Rico.
Unfortunately, years of development proved to be a costly undertaking, and after the company closed in 2021, Loon stopped work on LTA-HAPS. Other companies have taken up Mantle, continuing to develop their own LTA-HAPS for use in emergency communications, disaster recovery, providing private wireless networks, and extending offshore coverage.
A study by Machado and colleagues was published in the journal Nature, and in the video below, Machado discusses his work on developing autonomous navigation systems using reinforcement learning in more detail. This lecture is part of the AI seminar series offered by Amii Intelligence.
AI Seminar Series: Marlos C. Machado – Autonomous Navigation of Stratospheric Balloons Using RL (Jan 22)
Source: Nature
