Comprehensive Classroom Lecture Notes
Nonlinear Programming: Applications in Discrete Optimal
Control, Production-Inventory, Stochastic Resource Allocation,
and Facility Location
Prepared for Classroom Teaching
1. Discrete Optimal Control
Optimal control problems concern the determination of decision variables, called controls,
that guide the evolution of a system over time. In discrete optimal control, the planning
horizon is divided into finite periods k = 1,…,K. At each period, the system is represented
by a state vector y■, updated using control u■.
General Formulation:
State equation: y■ = y■■■ + φ(y■■■, u■)
Objective: Minimize A(y■,…,yK,u■,…,uK)
Constraints: y■ ∈ Y, u■ ∈ U, trajectory restrictions in D
Teaching Note: This structure matches nonlinear programming because both objective
and constraints may be nonlinear in y and u. Graphical representations (trajectories) are
useful in lectures.
2. Production-Inventory Example
This is a canonical example of discrete optimal control applied to operations management.
A company must meet demand across K periods with costs arising from labor changes
and inventory holding. The balance equations connect labor force, production, and
inventory.
Model Formulation:
Decision variables:
- L■: labor force at end of period k
- I■: inventory at end of period k
- u■: change in labor force in period k
Dynamics:
L■ = L■■■ + u■
I■ = I■■■ + pL■■■ – d■
Objective:
Minimize Σ (c■u■² + c■I■) for k=1,…,K
where c■ penalizes labor fluctuations and c■ penalizes inventory holding.
Constraints:
0 ≤ L■ ≤ capacity
I■ ≥ 0
Initial states (I■, L■) given
Teaching Notes: This problem demonstrates state-control interaction and quadratic
costs. In class, sketch demand curve vs. labor/inventory trajectory.
3. Stochastic Resource Allocation
In many problems, demand or resource availability is uncertain. Stochastic programming
introduces chance constraints to ensure a high probability of feasibility.
Illustrative Example:
Suppose demand x is random with Normal(µ, σ²). We require production p to satisfy
demand with at most 1% shortage probability:
P(p < x + 5) ≤ 0.01.
Deterministic equivalent:
p ≥ µ + z■.■■σ
where z■.■■ ≈ 2.33 is the 99% quantile.
Teaching Notes: Highlight the transformation from probabilistic constraints to
deterministic nonlinear inequalities. Emphasize managerial meaning: service levels vs.
risk of shortage.
4. Location of Facilities
Facility location models determine where to open facilities and how to allocate customers.
They capture trade-offs between fixed setup costs and variable assignment costs.
General Formulation:
Decision variables:
- y■ ∈ {0,1}: facility open indicator
- x■■ ∈ {0,1}: assignment of customer j to facility i
Objective:
Minimize Σ F■y■ + Σ C■■x■■
Constraints:
Σ■ x■■ = 1 for each j (each customer assigned)
x■■ ≤ y■ (assign only to open facilities)
Nonlinear Extensions: Include squared distances, congestion effects, or nonlinear cost
scaling. This leads to nonlinear programming formulations.
Teaching Notes: Instruct students to link this to real-world logistics, e.g., warehouse
placement, hospital service areas.
5. Summary and Key Takeaways
1 Discrete optimal control frames sequential decision-making over time with state
dynamics.
2 The production-inventory model illustrates quadratic costs for labor changes and linear
inventory costs.
3 Stochastic resource allocation incorporates uncertainty using chance constraints
transformed to deterministic form.
4 Facility location models integrate fixed and variable costs, extending naturally to
nonlinear cases.
5 All these examples highlight how nonlinear programming provides a unifying
mathematical framework.