001
C3: Consensus Complementarity Control
A robot balances and slides a tray with no grasp at all, pure contact. Why the underlying optimization hides 2140 mode combinations, and the ADMM split that turns them into five small problems solved in parallel.
002
Three ways to solve the same hard problem
Branch & bound versus ADMM on one toy problem, side by side. Where the work actually goes, and why anytime beats optimal when the deadline is hard.