Tipply AI today announces that it has successfully simulated unpowered superpressure balloons tracking and homing in on targets, as well as station keeping, using its proprietary machine learning models that predict local weather patterns.
The results bring Tipply AI a step closer to unlocking navigation using weather patterns, an entirely new capability for high-altitude balloons.
Simulation results
One target tracking mission tracked a ship travelling at 18 knots (33km/h) for two days. The balloon successfully stayed within 50km of the target 44.8% of the time at an average altitude of 18.4km.
A second target tracking simulation, this time 10 days long, stayed within 50km of the target 29% of the time at an average altitude of 18.6km.
A station keeping mission, intended to keep the balloon above a fixed point, stayed within a 50km radius of the target 73% of the time over a five-day period.
Finally, two simulated homing missions saw the balloon travel from its launch point to within 12km and 16km of a target in five days, both starting from 471km away.
How the models work
Tipply AI’s models work by moving the balloon towards winds that it predicts, based on forecasts made at a resolution of up to 1.5km, will blow it closer to the target location.
The models use past observations of the winds to correct the forecast in a local area and optimise the path to get and stay close to the target.
Every few minutes, a new trajectory is replanned, based on new observations and any change in the target location. The balloon can only control its rise and fall and is otherwise unpowered.
In real-world use, the models could run either locally on the balloon’s hardware or remotely from a control station.
What it unlocks
These results lay the groundwork for future growth and business development, including tests with the wide range of partners who have already expressed interest in Tipply AI’s technology.
In particular, the ability to launch stratospheric balloons capable of station keeping, at a cost of only thousands of dollars, could unlock capabilities that would previously have required launching a satellite, costing orders of magnitude more. Time to deploy would also drop from years to days.
Tipply AI believes that its models could ultimately be used in industries including reusable rocketry, sustainable stratospheric operations and defence.