Abstract:
The study of dynamic ambulance redeployment, also known as move-up or sys- tem status management, is the main concern of this investigation. Move-up is a practice of dynamically deciding stand-by locations for free ambulances in attempt to achieve quick response times. In the first part of the investigation, we study optimal move-up policies based on three small-scale Markov models to gain insights. The first Markov model considers one ambulance and aims to maximise the benefit of move-up for just the next call. The second Markov model still considers one ambulance, but aims to maximise the average benefit per unit time over an infinite horizon. The third Markov model extends the second Markov model by considering two ambulances. Numerical experiments are used to gain insights into optimal move- up policies based on the three models. In the second part of the investigation, we present three move-up models for realistic-sized problems. The first two of these models extend existing work by proposing a new simulation-based optimisation algorithm. The third move- up model is a new integer programming model which incorporates some of the insights obtained from the small-scale Markov models. Simulation-based numerical optimisation is employed to tune the model parameters and consequently, the model can also be viewed as an approximate dynamic programming model. Artificial call data generated for the city of Auckland, New Zealand, are used for computational experiments. We find that when move-up is performed appropriately, it can significantly improve the system performance. Moreover, the integer program proposed in this work gives the best performance.