Optimization Explained (Part 2)

By January 28, 2015 No Comments

Fleet Management Software

In the first part of this blog post, we began explaining the optimization process behind Pantonium’s transportation management system and how it works. In this post, we will continue this explanation by examining the main parameters used in the optimization process in greater detail.

To properly assess the parameters of Pantonium’s fleet management software, one must understand the Vehicle Routing Problem. The primary objective of optimization in vehicle fleet management is to service the highest number of trips while utilizing the least amount of resources for the day’s orders. Therefore, the goal is to find the most feasible routing solution at the lowest cost. This is commonly known as the vehicle routing problem. The vehicle routing problem is an optimization problem relating to servicing customers with a fleet of vehicles, which is the basis for how transportation companies operate. Solving the vehicle routing problem is achieved through Pantonium’s custom-built algorithm, which performs continual iterations and takes into account different penalties, which is the foundation used for the optimization process.

Iterations, in simplest form, are repetitions of a process. Due to the countless permutations when trying to solve the vehicle routing problem, Pantonium’s algorithm uses heuristics through different iterations to find feasible solutions when minimizing the overall cost. Heuristics are techniques used in computer science and mathematical optimization to determine the best approximate solution based on a specific timeframe. The Pantonium algorithm has been developed to utilize heuristics in order to create transport solutions optimized not just for overall planning, but for up-to-the-minute efficiency.

The number of iterations required to find a feasible solution can be dependent on the penalties set for the different parameters. This can include load time, ride time, duration, time penalty or variables affecting the trips such as additional trips, lateness, traffic, potential missed appointments, among others. In addition, generally the number of transports at a particular depot affect the complexity of the problem and hence the iterations required to find a feasible solution. This is different from the technology used by other market leaders who solve the vehicle routing problem in a pre-planned batch manner and do not continually run iterations. By continually running iterations with the software, Pantonium take into account real time changes in constraints when calculating a solution.

In addition to the optimization algorithm minimizing time and mileage, penalties associated with the parameters also play a role in the cost calculation. This allows the different parameters such as ride time to be also considered by the algorithm. Therefore, if the set parameters are violated when running through the different permutations and iteration process, a penalty is added for each violation, which in turn affects the end results. This means the penalties play an important factor in the cost calculation. Penalties add cost to do something, enabling a user to visualize and understand how certain parameter actions performed on fleet cost in relation to resource optimization.

This brings us to local optimals. A local optimal is the most optimum solution dependent on the neighbouring set of possible solutions presented (as opposed to a global optimal, which is the optimum solution among all possible solutions). The cost of a local optimal determine how desirable a particular solution is. There are two parts of the cost of a local optimal: the actual cost, and the optimization cost. The actual cost is based on the total time used by all the vehicles for the day to carry out the plan. The optimization cost takes into account the penalty cost which are occurrences which cost the company resources to accommodate. For example, a penalty can result from overscheduling the amount of trips for a single vehicle within a certain time frame. Some vehicles can only take a certain amount people, so if the vehicle is required to take more, penalties will be accrued for cost.

There are different types penalties used in the algorithm:

Load Penalty – This penalty is associated with the capacity of the vehicle. Generally this penalty is set high as vehicle capacities are relatively fixed.
Duration Penalty – This penalty is associated with the shift length and is based on the amount of times added for each time unit longer than the desired shift length of each vehicle.

Time Penalty – This is penalty is associated with pickup and appointment time windows and be the amount of time added for time unit a transport will be late for the scheduled time.

Ride Time Penalty – This penalty is associated with the time a rider can be on the vehicle. It is the amount of time to be added for each time until a rider stays on a vehicle longer than the maximum calculated ride time allowed for the transport.

The Pantonium platform provides a dashboard allowing for users to adjust the different penalties, which in turn affects the optimization results. As set parameters may be violated, the Pantonium system allows dispatchers to plan the most effective routes based on their knowledge and best practices. This way, dispatchers still retain full control and visibility of their operations while the Pantonium software is continually working to find the strongest solutions to the ever-changing vehicle routing problem.

For more information on how Pantonium’s fleet management software helps to optimize transportation management systems, please feel free to email us at or check out a demo of our system here.

Leave a Reply