Brutus is based on very delicate handling of background data and is able to create highly accurate predictions of daily transport flows in a chosen region. On this page the features are elaborated on for modeling experts wishing to take a bit more detailed look at the model.
In transport modeling (and usually also in reality) every trip is made for a purpose and it presents a certain utility for that person. In transport models, those utilities are presented as utility functions. Simulation in this context means that the estimated utility functions are applied stochastically using decision probabilities to solve single households’ and individuals’ decision making situations. By repeating these simulated decisions millions of times, it is possible to create the whole picture of the mobility patterns in the area, which corresponds to the reference distributions of mobility made available around the world by statistics offices or transport authorities.
Brutus is a stochastic, discrete Monte Carlo simulation model. The difference with the previously used models is, that Brutus deals with individuals rather than mean values within enumeration areas, trip chains rather than separate trips, and a 250 meter grid rather than large-scale enumeration areas. Brutus consists mainly of two models: a destination choice model and a mode choice model. The former is used to construct trip chains in each travel mode by choosing trip destinations one by one. The latter is used to make decisions between those alternatively constructed trip chains. Also a highly accurate assignment algorithm for cycling has been developed.
Being a highly accurate strategic simulation model Brutus demands a variety of data to work efficiently. The estimation of the model requires the following data
The present mobility patterns can be attained from travel diary-type studies. This kind of study also provides background data about households and individuals. The geocoded mobility data is connected to the land use data about the trip destinations, and this is used to describe the utility achieved by making a particular trip. Many authorities offer land use data about housing and residents, employment, schools, shopping facilities etc. and this data is used also in Brutus. In some countries the data is offered in an accurate 250 m x 250 m grid, and in some countries in post code areas or other divisions. The more accurate the data, the better is the outcome of the model estimation. As transport networks, it is possible to use for example network descriptions of traditional transport models. For cycling network it is possible to use OpenStreetMap data or any other accurate local data available. All network quality parameters are useful to improve the outcome of the assignment algorithm.
The travel resistance of the cycling network is described by setting a realistic (or measured) mean cycling speed to the actual cycling network, and a slower speed to the complementary network of less quality. This way the cycle trip assignment favors the actual cycle paths and lanes in a realistic way. Usually the following parameters are taken into account in the calculation of the travel resistance for cycling:
Hence, the value of travel resistance incorporates also comfort and safety issues experienced by the individuals. Also, if the amount of cycling is causing delays for other cyclists (“cycling congestion”), this effect can be taken into account.
First, the whole population of the area is constructed by replicating the households and individuals observed in the regional mobility study in a manner defined by certain explanatory variables until the population forecast is met. After this, the simulation proceeds as follows:
Thus, we assume that in the future there are similar households and individuals as they are today and they make as many trips to similar places as today. However, the chances in the transport networks and land use affect the destination and mode choice. As a result of the simulation Brutus gives every trip of every individual made in one day in the region. For every trip, Brutus produces trip origin and destination location and type, travel mode, departure time, travel time and time used in the destination.
There are 11 types of destinations in Brutus: own home, own work, other residential/visit place, work related visit, own school, “kiss&ride”, day care, shopping, restaurant, sport/culture/other free time place. There are five modes: walking, cycling, public transport, car driver and car passenger. New modes such as e-bikes are possible supplements in the future when more data about their use becomes available.
The destination choice of a single trip is created for each trip class and travel mode using alternative trip destinations’ land use and travel times.
With this definition the trips are distributed more probably to large land use than small, and closer than far away. However, “closer” here refers to the additional distance caused by one individual destination within the trip chain rather than the origin of the trip.
Mode choice is determined logically for the complete trip chain, so that one’s own vehicle (bicycle, car) is not abandoned during the trip chain. Nor can the vehicle be used from a location where it was not left earlier in the chain. Mode choice models are logit models, whose utility functions include travel times in the trip chain, the land use reached during the trip chain and various background data like household size, gender, age and employment.
The stochastic destination and mode choice models define choice probabilities that are used to set trip destinations and travel modes for each single trip. By repeating these stochastic choices one by one for each citizen, the result is a description of the trips of each citizen, presented with good geographical accuracy and with similar variation as in the actual mobility study. From these results it is possible to calculate statistical figures for the grid (250 m x 250 m), for any statistical areas or the whole region. These include: mode share, kilometers travelled, emissions etc.
The trips are assigned to the network, which enables analysis of trips on each link. The results can be analysed by each trip class, trip origin or destination and also trips going through particular cross-section in the network can be separated. The model results are validated using vehicle/cycle count data.