High altitude platforms (HAPs) such as stratospheric balloons and eventually other high altitude, long endurance unmanned vehicles have reached a stage where it is possible to deploy a persistent fleet of aircraft acting as a meteorological observation network for a reasonable cost. Whether directly collecting in situ measurements like winds aloft or via dropsondes or performing remote sensing using, for example, radar or GPS radio occultation, these observation networks can collect measurements which are hard to obtain from other observation platforms and are complementary to other systems. They are also highly autonomous and can be deployed worldwide (and thus can add redundancy to the global forecast system). Because they are mobile, the observation network can be adjusted to collect in situ measurements in the places that are most important to forecasters and scientists. We use simulation of fleets of stratospheric balloons that are navigated by machine learning algorithms that actuate an altitude control system to demonstrate some of the potential constellations that are achievable with HAPs and motivate the greater consideration of an autonomous, persistent HAPs-based meteorological observing network.