Food for Though: Considering Food Quality in Routing Procedures

Traditionally, vehicle routing optimization procedures focus on reducing cost by either minimizing distance or duration traveled. For perishable food products, however, one additionally may want to consider quality impacts in related routing and scheduling operations.

This leads to the question of how one can implement food quality aspects in such optimization procedures. Depending on the final use case, various decisions must be taken:

1. What means quality for my food product?

One of the beauty about working in the food sector is that each product has its unique characteristics and challenges. Consequently, before we can think about how one can integrate food quality aspects in routing and scheduling procedures, we need to define how we want – or even can – measure quality. For this, one can find various food quality models in literature, which often focus on the specifics of a single individual product or product class. In most cases, the result of this analysis is that time and temperature is of relevance, but there are plenty of other factors which one may want to consider such as aesthetics, food safety requirements as well as social, ethical and sustainability aspects.

2. What is my objective?

Once we have a clear picture on what quality means for our use case, we will need to think about how we can use it in our optimization models. There are mainly three ways on how one can integrate food quality aspects: (i) as a key part of the objective function; (ii) as a hard constraint that needs to be fulfilled, or (iii) as a soft constraint where violations result in penalties.

Adding food quality losses in the objective functions can be either done with a weighted objective function or in a multi-objective setting. The latter, however, makes your problem substantially more complex to solve and may lead to challenges when interpreting results. Nevertheless, as shown in multiple research studies, it can lead to highly relevant insights on how one can improve food quality.

Considering food quality loss as a constraint is more straightforward. In this case, a threshold value needs to be defined and food quality requirements are either fulfilled or not in the generated routing solutions. If modeled as a soft constraint, violations are additionally added to the objective function. The result of such an implementation, however, is that feasible solutions are not differentiated by how many food quality losses occur, but rather if the level of food quality loss is deemed acceptable or not.

3, How do I consider quality?

Now the final missing point is how generated vehicle routes are evaluated in respect to food quality. This point highly depends on your earlier decisions taken in Step 1 and 2. Let’s assume two different settings of increasing complexity:

• A minimum threshold (e.g., deadline) until when the products need to be delivered was set.

This setting will work well if your product has clear data marking (e.g., a best before date) and is not too susceptible to storage and handling conditions while in transport. Consequently, a latest delivery time for each product can be set and integrated in the vehicle routing procedure. The simplest way to do so is to use the deadline as a time window at the final delivery node, either as a hard or soft constraint as described above.

• A single time-dependent index was developed to approximate food quality losses over time.

The second setting considers that transport conditions additionally have an impact on food quality. This is of importance for products which are highly perishable or vulnerable to temperature changes. Common implementations assume that the products start at 100% quality. If it reaches zero, the product is lost and cannot be delivered successfully any longer. Consequently, the respective vehicle tour is infeasible. How the quality decreases over time again depends on the products characteristics and design choices taken before. Common considerations are to deduct quality percentage points for time spend before pickup and while in transit. Additionally, loading and unloading operations are often critical due to door openings. One might want to include various additional factors here such as product mixes, environmental conditions (e.g., weather effects, traffic levels) or even the driving behavior of the individual food delivery trucks.

Answering these questions early during your research or development project will help you to design an efficient optimization procedure and save you a lot of time during later stages. If you are interested further in this topic, I provided some references below which show successful research examples of how such a consideration of food quality aspects can work in related routing procedures. Please let me know if you know of additional literature or have further experiences and hints to share one how to integrate food quality losses in vehicle routing and scheduling procedures.

References

Image: MVOA