When designing complex machinery like spacecraft or drones, engineers typically use best- and worst-case scenario thought processes — kind of like what’s the best we can get out of this part and what’s the minimum this part can do?
This kind of thinking, however, doesn’t account for things that happen in the real world, like battery life depleting faster in the heat or software behaving differently in different conditions.
For this reason, a team from MIT and the University of Zurich created a framework that allows for probability. So instead of factoring in that a certain motor will last five hours, the framework says sometimes it lasts five hours, sometimes it lasts four and at times it lasts six hours.
Each part comes now with a probability profile that allows the designers to run “what if” simulations that factor in the messiness of the real world.
Testing was run on drone design and the results were unexpected. For example, a specific battery setup appeared poor in worst-case models but performed very well on average. Without the new probability-based considerations, it may have been dismissed altogether.
This means engineers can design smarter, safer and more dependable systems, which is imperative in situations where failure is not an option. Embracing uncertainty has made design more realistic and much more of service.
The research team will be speaking on their paper at the IEEE Conference on Decision and Control (CDC) 2025.
More like this: Croptimal choices
