Probabilistic reasoning for meal planning in intelligent fridges

In this paper, we investigate issues in building an intelligent fridge which can help a family to plan meals based on each member's preference and to generate a list for grocery shopping. The brute-force solution for this problem is intractable. We present the use of a BNF grammar to reduce the search space. We select the meal plan from alternatives following a decision-theoretic approach. The utility of a meal plan is evaluated by aggregating the utilities of meals and foods contained in meals. We propose an explicit representation of the uncertainty of each family member's food preference using extended Bayesian networks.