Generalized Hypercube Queuing Models with Overlapping Service Regions

We present a generalized hypercube queuing model, building upon the original model by Larson 1974, focusing on its application to overlapping service regions such as police beats. To design a service region, we need to capture the workload and police car operation, a type of mobile server. The traditional hypercube queuing model excels in capturing systems' dynamics with light traffic, as it primarily considers whether each server is busy or idle. However, contemporary service operations often experience saturation, in which each server in the system can only process a subset of calls and a queue in front of each server is allowed. Hence, the simple binary status for each server becomes inadequate, prompting the need for a more intricate state space analysis. Our proposed model addresses this problem using a Markov model with a large state space represented by non-negative integer-valued vectors. By leveraging the sparsity structure of the transition matrix, where transitions occur between states whose vectors differ by one in the ℓ_1 distance, we can solve the steady-state distribution of states efficiently. This solution can then be used to evaluate general performance metrics for the service system. We validate our model's effectiveness through simulations of various artificial service systems. We also apply our model to the Atlanta police operation system, which faces challenges such as increased workload, significant staff shortages, and the impact of boundary effects among crime incidents. Using real 911 calls-for-service data, our analysis indicates that the police operations system with permitted overlapping patrols can significantly mitigate these problems, leading to more effective police force deployment.