liveBook · Manning

Production planning

As explained in Applied Reinforcement Learning

Production planning is the process of deciding what products to manufacture, in what quantities, and when, subject to capacity, demand, and cost considerations.

A factory must choose what to make each day while balancing limited machine capacity, uncertain demand, setup changes, and inventory costs. A Markov decision process represents the decision over a fixed planning horizon. section 2.5

Model the daily decision with:

  • A state containing the day, inventory of each product, and current machine setup. section 2.5
  • Actions to produce scooters, produce bikes, or idle. section 2.5
  • Stochastic demand, inventory transitions, and rewards that account for sales, production, setup, holding, and backorder costs. section 2.5

Two-product factory

A scooter-and-bike plant can use greedy or setup-aware policies, then simulate each policy over the planning horizon and track cumulative profit. section 2.5

Listing 2.4 Factory problem definition
Class FactoryProblemDefinition:
    lambda_scooter = 4.0 #A
    lambda_bike = 3.0 #A
    horizon = 30   #A
 
    prod_capacity_scooter = 8 #B
    prod_capacity_bike = 6 #B
 
    price_scooter = 40.0 #C
    price_bike = 50.0 #C
    cost_scooter = 20.0  #C
    cost_bike = 25.0  #C
 
    setup_cost = 200.0    #D
    holding_cost = 2.0    #D
    backorder_cost = 10.0 #D

Factory parameters specify demand rates, production capacities, prices, production costs, setup costs, and inventory penalties. section 2.5

How Applied Reinforcement Learning covers it

How deep it goes
Defines states, actions, transitions, rewards, heuristic policies, and trajectory simulation for a two-product factory.
Examples
Uses scooter-and-bike production, setup changes, inventory costs, and greedy or setup-aware policies.
What you can do after
Build it from scratch: encode factory states, production actions, stochastic demand, rewards, and policy simulation.
How it connects
Links Markov decision processes to inventory, stochastic demand, policies, transitions, and reward functions.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Applied Reinforcement Learning. The section links point to the parts of the book it draws on.