When a dynamic programming problem becomes too large for an exact solution, approximation sacrifices some exactness to keep computation tractable. The formulation may remain valid even when its state space exceeds practical computational limits. section 6.7.1
State-space growth can come from adding independent constraints or representing a variable in finer detail. For example, tracking inventory in single-unit increments creates more states than tracking it in increments of ten. Aggregating states is one way to reduce the burden, though it gives up some exactness. section 6.7.1

