[gerthworm]

We took the N'th degree approach this year, and it turned out relatively well:

-Monitor battery condition over time. Generate an adaptive model of the battery to determine how much current can be drawn from it before system voltage drops too low.

-Proactively estimate how much current draw a given driver command will pull from the battery. Involves knowledge of the physical parameters of the motor, and its present speed (measured via encoders)

-If the estimated current from the driver commands is sufficient to cause the system voltage to go too low, scale back driver commands.

The algorithm we used this year is still in trial - it kicked in plenty during competition, and we never browned out. However, we don't have good enough data yet to support that it was 100% necessary. We are working on a whitepaper (there are a few drafts of it floating around the internet), and will hopefully by end of summer have more data on the limits of what we came up with.