Group Activity Planner (GAP)

Summary

GAP is an AI program (agent) that provides assistance to any team or group that is attempting to schedule an activity. At a high level, the agent will increase its knowledge about the user and their preferences. Given the knowledge base that it creates, it can help its users to schedule activities. A user can request a specific date and a specific time that he/she would like a group of users to meet for an activity. The agent is very useful in providing users with scheduling capability based upon user preferences/profiles, and their current schedules.

GAP is composed of objects including the agent, the users, user preferences, user schedules, and activities. In the future, GAP can be enhanced to have properties on activities to add even more complexity by associating properties such as location (e.g. hold an activity in California… even though a schedule may permit it, travel time/expense may not), or physical constraints (e.g. if an activity is sky diving, certain people with certain medical conditions may not be able to participate even though the schedule may permit it). GAP can grow to include much more knowledge (as shown by adding properties to Activities), however, for the purposes of this project, GAP will make the following assumptions:

• If the schedule indicates that the date/time slot is available for a particular user, then the activity (regardless of nature) can be added to that user GAP will only be aware of activities that the user has notified it of (e.g. the user may have scheduled a meeting and not let the agent know… in which case according to the agent, the user is free during that date/time interval).
• Activities are not differentiated.
• Preference is always given to earlier date/time slots (in the case where the agent is requested to find a date/time slot) that a group can meet. A group is 1 or more users.
• All users are defined by name and all users have a preference file (which may be blank) and all users have a schedule (which also may be empty). A user name is a single word and two users cannot have the same name.
• All time slots are one hour in duration and military time (multiples of one hour can be requested).

Illustrative Example

1. We have to let the agent know who are all the users.
   • We have five users: Tom, Mary, John, Dave, and Lisa.
     • add_user('Tom').
     • add_user('Mary').
     • add_user('John').
     • add_user('Dave').
     • add_user('Lisa').
   
   
2. Then we have to give preferences for the users. This is the time that a user will not be available for a meeting.
   • Mary does not want to see anyone on 11/28/99 from 8:00 am to 12:00 pm.
     • add_preference('Mary', 11/28/99, '8:00', '12:00').
   • John want's to go to golf on 11/29/99 from 1:00 pm to 3:00 pm.
     
     • add_preference('John', 11/29/99, '13:00', '15:00').
   
   
3. Now the user can try to schedule meetings:
   • Tom's wants to have a meeting to discuss operating systems with Mary and Dave on 11/28/99 from 10:00 am to 1:00 pm.
     • schedule_meeting('Operating Systems', ['Tom','Mary','Dave'], 11/28/99, '10:00', '13:00').
     Result:
     [ Could not schedule meeting ]
     Users available: Tom Dave
     Users unavailable: Mary
   • Lisa wants to get together with Mary, Dave, and John to study aritificial intelligence on 11/28/99 from 6:00 pm to 9:00 pm.
     • schedule_meeting('Artificial Intelligence', ['Lisa','Mary','Dave','John'], 11/28/99, '18:00', '21:00').
     Result:
     [ Meeting scheduled for: Lisa Mary Dave John ]
   
   
4. Now the user can see their schedules.
   • Lisa wants to see her schedule.
     • show_schedule('Lisa').
     Result:
     Artificial Intelligence 11/28/99 18:00 21:00
   
   
5. The agent can print the schedules and preferences of all the users.
   • print_all.
   Result:
   **************
   User: Dave
   Preference:
   Schedule: Artificial Intelligence 11/28/99 18:00 21:00
   **************
   **************
   User: John
   Preference: 11/29/99 13:00 15:00
   Schedule: Artificial Intelligence 11/28/99 18:00 21:00
   **************
   **************
   User: Lisa
   Preference:
   Schedule: Artificial Intelligence 11/28/99 18:00 21:00
   **************
   **************
   User: Mary
   Preference: 11/28/99 8:00 12:00
   Schedule: Artificial Intelligence 11/28/99 18:00 21:00
   **************
   **************
   User: Tom
   Preference:
   Schedule:
   **************
   

TECHNIQUES

There were three main challenging areas of development of the GAP application:

1. Searching the Tree,

2. Saving State Space changes after each command, and

3. Conflict Checking.

The GAP agent saves the state space in a tree format. Each node on the tree consists of the user, preference, and schedule for each individual. Each state is traversed using the depth first search mechanism. The agent uses depth first search to both find and to update the knowledge base database about users, preferences and activities that are being scheduled.

After the user enters a command, the change in the state space is reflected before the next command. This is similiar to the situation calculus discussed in class. The new situation (environment) is passed on to the next command, thus saving the state space changes after each command.

The agent also determines if there are conflicts with schedules and preferences before scheduling any meetings. It performs this conflict checking by using constraint satisfaction techniques. Knowledge about existing schedules and preferences are stored in domain's knowledge base. Rules are created that use this knowledge to constrain their firing; once the constraints are satisfied, the rules can be executed.