DEDALO provides a framework to design an informed proposal distribution in the context of Metropolis-Hastings algorithms employed to find the agent-based stochastic user equilibrium in MATSim . The proposal distribution is used to select new travel choices (e.g., transport mode between two activities, departure time, routes) in an activity-based model, which represents the daily plan of every agent in the synthetic population. In order to obtain a better convergence rate and reduce the auto-correlation between samples , the proposal distribution needs to approximate the target distribution which is considered, in this work, a mixture-logit model. However, to be computationally efficient, this requires to solve two main obstacles. The first one stems from the combinatorial nature of all the possible choices a traveler deals with during a daily plan and the second one is due to the unknown stochastic disturbance the agent faces during the execution of his/her plan as a result of the stochastic decisions of all the other agents. DEDALO overcomes these obstacles, providing the agents with an augmented capacity to forecast the choices of the other agents by knowing the probability of their decisions and adapting accordingly.
- JRE 11: Java runtime version 11 is needed for using the library.
- JDK 11: The Java JDK 11 must be installed.