For the problem sizes the author looks into, MILP would do fine. Plus you would get optimality guarantees (or at least a meaningful optimality gap) that a trill

For the problem sizes the author looks into, MILP would do fine. Plus you would get optimality guarantees (or at least a meaningful optimality gap) that a trillion agents will never give you.