Eyes in the Sky: Distributed and Autonomous Earth Observation

Perhaps the most important result of human space exploration is that it has caused us to better appreciate our own planet. But despite massive progress in recent years, Earth-observing satellites still have significant limitations. Recent technological developments, though, are reducing satellites’ costs and are also improving their ability to coordinate with each other. To learn about how this new generation of satellites will help us to better understand Earth, we spoke to Daniel Selva, an aerospace engineering professor at Texas A&M University who specializes in space systems. Dr. Selva furthermore directs the Systems Engineering, Architecture, and Knowledge (SEAK) Lab.


What are the current shortcomings of systems of Earth-observing satellites?

Perhaps the most important shortcoming is that the kind of Earth-observing satellites that NASA and ESA are developing are still quite expensive – on the order of hundreds of millions of dollars, often more than a billion. This is mostly due to two things: 1) access to space is very costly, on the order of $10,000 per kilogram of satellite; and 2) the space industry is extremely risk-averse and tries to avoid failures at any cost. These two facts reinforce one another in what some people refer to as the space “death spiral”: getting satellites to space is expensive so we want to make them highly reliable, which in turn reduces the number of organizations that can afford building satellites, which in turn decreases the demand for launch services, which feeds back into further increasing prices. It’s a vicious cycle.

Another important shortcoming of Earth-observing satellites today is that for the most part they are not very “smart”. All they do is take images of Earth with various instruments, store those images on board, wait to be visible to the next ground station, and then dump as much data as possible during contact with the ground. Very little data processing is actually done onboard. All the intelligent planning and scheduling is done from the ground. This makes sense from a cost perspective: it is more expensive to put intelligence onboard than it is to put it on the ground. However, this has implications for the operational agility of the satellites because it means that the ground stations end up relying on data that can be quite stale; most satellites only have a few contacts with the ground per day.

What are some ways you think these shortcomings can be addressed?

The first way is by having more distributed systems. We are undergoing an architectural paradigm change in satellite Earth observation – from large “monolithic” systems carrying multiple instruments to more distributed systems consisting of many small satellites (e.g., Planet). This leads to many advantages: affordability, since each satellite is cheaper to develop and the system can be deployed over time; shorter development times, which means lower risk of obsolescence; resilience, since the system can still operate in the event of a satellite or launch failure; and in some cases, increased temporal, spatial, or angular sampling due to the diversity of measurements.

The second way we can improve the shortcomings of Earth observation satellites is by building in more autonomy. The combination of onboard data processing and inter-satellite links can enable Earth-observing systems to overcome the operational agility problem. For example, a constellation of satellites tracking hurricanes can run storm detection algorithms onboard, and upon detection of a storm, can then automatically tune the parameters of their onboard instruments to improve the sensitivity of those instruments for that particular storm (e.g., channel selection). Satellites could also then communicate with other satellites in their constellation about the storm, which could cause them to decide to point and tune their payloads to that particular storm as well. And, importantly, all of that would happen autonomously, without any input from ground control.

The two factors are synergistic. As we develop larger systems consisting of hundreds of satellites, our current approach to operating spacecraft – essentially one person per satellite – will not scale. Increased spacecraft autonomy will thus be an enabler for the growth of distributed space systems.

What motivates you to study Earth observation systems?

I think it is a combination of things. There is something vocational or even spiritual about pursuing space exploration that has always attracted me. But at the same time, satellite Earth observation has very critical and pragmatic uses for society – satellites play an important role in weather forecasting, disaster monitoring, and climate monitoring, for instance. And then there are the fascinating engineering challenges behind space systems in general and Earth observation satellites in particular. The first time I heard, for instance, that a satellite consuming only a few hundred Watts can measure soil moisture on the entire world to within 4%… I just needed to learn how that was possible. And that’s how it all started for me.