Hospital Instances

Hospital Instances [as described in Jungwirth et al., "The vehicle routing problem with time windows, flexible service locations and time-dependent location capacity, Technical Report, 2020] Network Layout: We distinguish three layouts, which are defined by the number of buildings B ∈ {1, 2, 6} and the number of floors per building F ∈ {1, 3, 6}. Each floor has a certain number of rooms (locations) R ∈ {6, 7, . . . , 10} drawn from a discrete uniform distribution. One of the buildings contains the therapy centers, which has a capacity between 2 and 6. The capacity at the ward rooms is always unlimited since patients cannot be scheduled to other patients’ ward room. The travel time between two buildings is drawn at random from the set {10, 15, 20} minutes. The travel time between neighboring floors is assumed to be 5 minutes, and the travel time between two rooms on the same floor is either 5 or 10 minutes. Demand Scenarios: We distinguish six demand scenarios having 20, 40, 60, 80, 100 and 120 treatments. A 10% probability exists that the patient is an outpatient, i.e. he/she can only be treated in a therapy center and the start time for the treatment is fixed. Ten percent of the patients are bedridden, i.e. the patient must not be moved and can only be treated in his/her room at the ward. However, these latter patients have a rather wide start time window of 90 minutes. The remaining patients are regular inpatients, of which 50% have location flexibility, i.e. the patient can be treated at the ward and in the therapy centers; however, a preference for one location exists, generally the ward room. The start time window length of these patients varies between 30 and 45 minutes. Thirty percent of the patients receive multiple treatments (2 or 3) in one day. Every treatment job has a duration of 10 to 45 minutes and requires a certain skill level. We assume hierarchical skills ranging from 1 (lowest) to 3 (highest). The probabilities that a job requires a certain skill are 60%, 30% and 10% for skills 1, 2 and 3, respectively. Vehicles: We use a heterogeneous fleet, since therapists differ in their skills as well as their shift patterns. The skills are the same as for the jobs; however, the probabilities of having skill 1, 2 and 3 are 10%, 60% and 30%. A therapist has a regular (long) shift with 80% probability. Otherwise, the therapist has a short shift with 50% probability of being a morning or evening shift. We assume that therapists start and end their shi