What To Pack For Mars

A successful mission requires a well-planned supply strategy

2 min read
Sketch of Olivier de Weck
Illustration: Jacob Thomas

This is part of IEEE Spectrum’s Special Report: Why Mars? Why Now?

Sketch of Olivier de Weck Illustration: Jacob Thomas

You can’t take much into space. When a rocket is on the launchpad, 99.9 percent of the mass is the fuel and the vehicle itself. That leaves 0.1 percent for everything else—the crew and all their supplies. When considering what to bring, how do we trade off between consumables (needed for survival), spare parts (for safety), and research equipment (which gives the mission value)?

Starting in 2005, NASA asked my group at MIT to develop SpaceNet, software that helps mission planners evaluate these trade-offs. The program manages and models the complex supply chain of vehicles and supplies along with the processes and orbital dynamics required for manned missions, whether they’re to Mars, the International Space Station (ISS), or a lunar outpost. We designed our software to model each step in a mission as well as a whole campaign of missions. SpaceNet allows planners to quantitatively compare different mission architectures to optimize the exploration capability and launch mass. We also want to make supply chains robust so that one failed or delayed mission doesn’t ruin the whole plan.

In 2005, we tested our computer models by participating in the Haughton-Mars Project, in which a small group of researchers live in an Arctic base as if they were on Mars. The experience was enlightening. We found that of the operational inventory (that 0.1 percent of launch mass), two-thirds went to ground vehicles and fuel for powering the base. From this experience, we’ve calculated that each crew member added to a 600-day Mars mission would require sending 13 metric tons more cargo to the Martian surface.

But even if you deliver the right amount of supplies, it matters how they are organized. For example, there are between 15 000 and 20 000 objects on the ISS. If you take the total number of useful crew hours in a year and divide by the total operating budget, you find that the value of 1 hour of an astronaut’s time on the ISS is US $186 000. So 5 minutes spent looking for one hard-to-find item wastes $15 000.

On a Mars mission, time will be even more valuable. Storage should be reconfigurable so that the most needed items are always accessible and everything else is out of the way. We’re now working on an RFID system that tracks the location of each piece of inventory at all times. The ultimate goal is to create smart, self-aware environments that are both safe and effective for exploration far from Earth. We hope that better space logistics will give future astronauts more time to do valuable work.

—As told to Joshua J. Romero

For more articles, go to Special Report: Why Mars? Why Now?

About the Author

Olivier L. de Weck is the leader of the MIT Strategic Engineering Research Group.

The Conversation (0)

How Can We Get​ Blockchains to Talk to Each Other?

The field is fragmented, but common protocols are on the way

4 min read
An artists impression of two blockchain blocks trying to communicate with each other.

This article is part of our exclusive IEEE Journal Watch series in partnership with IEEE Xplore.

Keep Reading ↓ Show less

Clever Compression of Some Neural Nets Improves Performance

MIT researchers find an efficient way to prune speech-recognition AIs while still boosting accuracy

3 min read
Icon based illustration showing an AI language processing cycle.

As neural networks grow larger, they become more powerful, but also more power-hungry, gobbling electricity, time, and computer memory. Researchers have explored ways to lighten the load, especially for deployment on mobile devices. One compression method is called pruning—deleting the weakest links. New research proposes a novel way to prune speech-recognition models, making the pruning process more efficient while also rendering the compressed model more accurate.

The researchers addressed speech recognition for relatively uncommon languages. To learn speech recognition using only supervised learning, software requires a lot of existing audio-text pairings, which are in short supply for some languages. A popular method called self-supervised learning gets around the problem. In self-supervised learning, a model finds patterns in data without any labels—such as “dog” on a dog image. Artificial intelligence can then build on these patterns and learn more focused tasks using supervised learning on minimal data, a process called fine-tuning.

Keep Reading ↓ Show less

Fundamentals of radome and bumper measurements using the R&S QAR

This whitepaper examines how the R&S QAR quality automotive radome tester can help

1 min read

Due to design challenges, radar sensors are usually hidden behind bumpers or emblems, which can influence how well the radar sensor can interpret the outgoing and incoming data. To test radome and bumper influence on radar sensors, this whitepaper examines how the R&S®QAR quality automotive radome tester can help. Register now and download our free whitepaper.