Academic Advisor: Prof. Ronen Brafman Team Members: Ran Isenberg Mirit Markovich Noa Aharon Alon Furman

System Requirements System Architecture Class Diagram User Interface Algorithm Task List

System Requirements Functional requirements Answering a question. Evaluating an activity Changing the duration of an activity in the schedule. Removing an activity from the schedule. Search for an activity Details about an activity Change city. Functional requirements Answering a question. Evaluating an activity Changing the duration of an activity in the schedule. Removing an activity from the schedule. Search for an activity Details about an activity Change city. Non functional requirements Speed Capacity Modularity Programming Language External API SE Project constraints Non functional requirements Speed Capacity Modularity Programming Language External API SE Project constraints

System Architecture Database

System Architecture (Cont.)

Class Diagram SCUFacade CityModel Handler CityModel Handler Solver GUI Controller GUI Controller Database DBHandler Exhaustive Solver Exhaustive Solver Distributed Solver Distributed Solver Schedule Activity

MCVMV Pattern View ViewModel Controller Model

User Interface Change City Choose city: HAIFA TEL-AVIV HADERA … Submit

The Algorithm The algorithm is based on the article “Asynchronous Forward Bounding for Distributed COPs, Amir Gershman, Amnon Meisels and Roie Zivan – '09”. Each agent represents an activity in a city. The domain values of each agent represent the time interval between the opening hour and the closing hour of the activity, partitioned into 30 min segments. In addition, the value of the best schedule found so far, and the number of agents who didn't participate in that schedule are stored as global variables.

Bahai Gardens Museum of Art German Colony Mall Schedule The first agent tries to assign a value from its domain. The agent sends the partial assignment to the next agent The agent also sends a request to all agents with id greater than his to estimate if they will participate in the assignment in order to calculate how many agents are not going to participate in the current assignment Meanwhile the second agent tries to assign a value from its domain, sends the partial assignment etc…

When some agent receives an evaluation from other agents regarding how many of them will not participate in the new assignment, he checks whether this value is larger than the number of agents not participating in the schedule found so far. If it is, the new schedule can’t be better than the one found already and the current branch can be pruned. If the last agent fills the partial assignment and the value of that assignment is greater than the current solution, he will advertise the solution to all agents.

Task List Development 15/02 – 30/04 Database & City Model diagrams 15/02 – 05/03 Persistence layer handlers 01/03 – 10/03 Solver 15/03 – 30/04 Logic Layer (Façade, GUI Controller) 30/03 – 20/04 Views and ViewModels 07/04 – 15/04 Testing + Unit Testing 01/04 – 30/04 Prototype Submission 30/04