Estimation of Waste Collection Vehicles' Travel Time Using Spatial-Temporal Artificial Neural Networks Algorithms (Case Study: Tehran City)

Travel time data is important for most of the vehicle routing problems (VRP) aimed at modeling time constraints like customer time window as well as time objectives, such as minimizing travel time and waiting time. One of the most important advantages of precise travel times in VRPs is the reduction of total travel time and increasing route reliability, which with more accurate travel time data better removes time constraints. The obvious reasons for not using variable travel time in VRP are related to three factors. First, conventional VRP algorithms cannot consider variable travel time without major structural changes. Second, all parameters affecting travel time are often not used. Third, collecting the required data is very difficult. Therefore, designing appropriate models with enough accuracy in predicting travel time based on the effective variables is a necessity in the transportation planning process.In this paper, the appropriate variables for measuring the impact of spatial-temporal and traffic factors on the travel time of medical waste collection vehicles in a single route were defined and by examining and implementing the proposed model, the impact of these variables on the travel time was measured. To estimate the travel time, we used the proposed model called spatial-temporal neural network (ST-NN) which can examine the effect of variables at any moment of the travel time. The study area is Tehran metropolis with its entire urban passages network. The required data were first pre-processed and then entered into the neural network algorithm as model inputs. The model was