Stochastics, CSE, Analysis, Optimization

Stokhos ( and youtube: stokhos) developed and markets an On-Time Emergency Response Solution (OTERS), which enable emergency services to minimize response times. This also improves the efficiency of the vehicle fleet and as such saves capital and operational costs.

Although we currently operate in the EMS (Medical Services), we will eventually branch out into other complex logistics market places, police, emergency elderly care, taxi’s etc.
Stokhos deals with a number of scientific challenges:

  1. How to develop a model to accurately predict the demand for ambulance care in both time and space?

In this context, the scientific complexity lies in the omnipresence of uncertainty in the demand process, for example in the inherent randomness in both the numbers of incidents requiring emergency care and their locations. To address this, we will use a wealth of historical data about past incidents and datasets about external factors like local weather circumstances and local traffic situations made available for this project, and to develop new forecasting models that can predict number of incidents and their locations in time and space. For this, we use analytics techniques in active machine learning and regression models.

  1. How to optimally route vehicles in case of for proactive relocations?

Algorithms for proactive relocations typically propose an ambulance movement form location A to location B but not which route to take, whereas the choice of the route from A to B may have a significant impact on the ‘coverage’ of the service area during the relocation. In this context the scientific challenge consists of two parts: (1) identification of a route corridor, consisting of a limited set of possible routes from A to B, and (2) selecting the specific route within the route corridor that optimally balances coverage added by the ambulance moving from A to B on the one hand, and the travel route from A to B on the other hand. To address the first question, we will use the powerful theory of so-called discrete-choice models, and the second part, we will use methods form stochastic optimization.

  1. How to modify the relocation algorithms for applicability to different countries?  

In general, the protocols and business processes are country-specific. The scientific challenge is two-fold: (1) the development of new relocation models tailored to specific country-specific business processes, and (2) the adaptation of the solution methods to incorporate the country-specific dynamics.

For example, in the Netherlands an ambulance physician usually travels to an emergency scene in the ambulance vehicle, whereas in Germany, the ambulance vehicle and the physician travel to the emergency scene independently. Consequently, in the modeling of the German system, the state space not only include the locations and statuses of the ambulance vehicles, but also of the physicians.

This seemingly simple modification puts a tremendous burden on the computational complexity of the optimal relocation algorithm, and hence, raises the need for developing smart and scalable heuristics to deal with the added complexity of the German system.

Go back to the MCOD homepage


Login with your CH account or TU Delft NetID.

Login with CH Connect

Having trouble logging in? Click here