A method for Constrained optimization, which generalizes Lagrange multipliers to Inequality constraints, rather than just equality ones

See here

https://www.wikiwand.com/en/Karush%E2%80%93Kuhn%E2%80%93Tucker_conditions

Imagine we add Lagrange multipliers but with conditions that must be greater than or equal to 0 (w.l.o.g.), then at the optimum these conditions are either greater than zero, in which case the multiplier is 0, or are 0 in which case the multiplier is non-zero, and enforces the solution to stay within the condition=0 contour.