Assessing Values-based Sourcing Strategies in Regional Food Supply Networks : An Agent-based Approach
agent then uses this information to determine whether or not to allow new producer agents to become members of the hub .
Model Overview
The producer and consumer agents trade six different categories of food , using the food hub as an intermediary . Each producer agent produces and sells one of the six product categories to consumers through the food hub . The categories and percentage of producers supplying them were : meat ( 25 %), dairy ( 5 %), eggs ( 9 %), fresh produce ( 36 %), ingredients ( 3 %), and processed convenience foods ( 22 %). Each time the model generates a producer agent , there is a fixed probability that the agent will be assigned to particular category ( e . g ., there is a 25 % chance that it will be a meat producer ), based on historical data from a real-life food hub . It is assumed that a producer agent may only provide items in a single category , which is typically true in the real regional food system ( Krejci et al ., 2016 ).
Each simulated time step represents a distribution cycle by the food hub , which occurs approximately every two weeks throughout the year , for a total of 22 cycles per year . Producers and consumers can be in one of three different membership states with respect to the food hub : nonmember , member , or canceled member . Agent interactions are confined to producer – consumer transactions . It is assumed that consumers do not interact with one another directly , and neither do producers .
The model consists of five major submodels : initialization , consumer purchase decisions , consumer evaluation and status update , producer evaluation and status update , and food hub membership update . The initialization submodel is only run once , at the start of each simulation run . The other four submodels are executed sequentially in every time step .
Initialization . In each simulation run , the model is initialized with 30 producer agents , each of which is randomly assigned parameter values based on the probabilities determined from the interview data , system data , and assumptions . Different random number streams and seeds are used for each run , such that the outputs of each run are statistically independent . Each producer is initialized with 100 % of its yield available for sale through the food hub . Fifty consumer agents are created , each of which is randomly assigned a demand category ( i . e ., low , medium , high ), a food familiarity category , and a persona . Each consumer ’ s producer rating matrix is initialized with producer attribute values for each of the producer agents in the model . A consumer ’ s overall utility is initialized to 1.00 ( the maximum value ), and food hub membership status for all consumers and producers is set to “ member .”
Consumer purchase decisions . Each consumer who is currently a food hub
33