Household Activity Pattern Problem Paper by W W
Household Activity Pattern Problem Paper by: W. W. Recker. Presented by: Jeremiah Jilk May 26, 2004 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004
Overview l General Concepts l Starchild, HAPP and PDPTW l 5 Cases l Conclusion Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 2
General Concepts l Activity Problem l l l “There is a general consensus that the demand for travel is derived from a need or desire to participate in activities that are spatially distributed over the geographic landscape. ” In other words, we travel because we need or want to do things that are not all in the same place. Spatial and Temporal l l Travel and Activities can be represented by a continuous path in the spatial and temporal dimensions. This is a simple concept, but is very difficult to implement operationally. Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 3
Starchild, HAPP and PDPTW l Starchild Model l l Best previous model Problems: § § l Model members of the household separately Exhaustive enumeration and evaluation of all possible solutions Discretizes temporal decisions Does not consider vehicle or activity allocation HAPP – Household Activity Pattern Problem l l The Goal of HAPP is to create a travel schedule of a household that accomplishes a set of activities. Avoid the problems of Starchild. Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 4
Starchild, HAPP and PDPTW l PDPTW – Pickup and Delivery Problem with Time Windows l l l Well known Problem of scheduling pickups and deliveries. Optimizes a utility function to get a set of interrelated paths for pickup and deliveries though the time and space continuum. HAPP – Household Activity Pattern Problem l l HAPP can be viewed as a modified version of PDPTW and can use the same algorithms for solving. Optimize a utility function to get interrelated paths through the time and space continuum of a series of household members with a prescribed activity agenda and a stable of vehicles and ridesharing options. Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 5
HAPP - Input Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 6
HAPP - Input Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 7
HAPP - Input Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 8
HAPP – Case 1 l l l Each member of the household has exclusive unrestricted use of a vehicle Any activity can be completed by any member of the household PDPTW l The demand function and vehicle capacity are important to PDPTW. They are unimportant to HAPP, but can redefined as follows: Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 9
HAPP – Case 1 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 10
HAPP – Case 1 l Disutility function (Z) l By minimizing the disutility function, we are optimizing the schedule. There are many disutility functions to choose from. The basic components of the disutility function are: Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 11
HAPP – Case 1 l Constraint Functions l Disutility Function l If u is an activity location, then there is a trip from u to some w l There are the same number of trips as back trips Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 12
HAPP – Case 1 l Constraint Functions l Vehicle v will travel to at least 1 activity l Vehicle v will return home l If v travels from w to u it will also travel to the return destination of u Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 13
HAPP – Case 1 l Constraint Functions l The time u starts + the time it takes to do activity u + the time it takes to get from u back home ≤ the time v gets home l If v goes from u to w, then the time u starts + the time it takes to do activity u + the time it takes to get from u to w ≤ the start time of w l If u is the first stop for vehicle v, then the start time + the time it takes to get from home to u ≤ the start time of u Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 14
HAPP – Case 1 l Constraint Functions l If v goes from u to the end, then the start time of u + the time it takes to do activity u + the time to travel from u to home ≤ the end time l The start time of u is within bounds l The start time for v is within bounds Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 15
HAPP – Case 1 l Constraint Functions l The finish time for vehicle v is within bounds l Moving onto another activity costs demand l Returning from an activity relieves demand Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 16
HAPP – Case 1 l Constraint Functions l Moving from home to an activity costs demand l Demand starts at 0 can not be less than 0 and can not be more than D l Vehicle v either goes from u to w or not Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 17
HAPP – Case 1 l Constraint Functions l The total cost of all trips can not be more than the budgeted cost l The total time vehicle v is on trips can not be more than the budgeted time l Vehicle v can not go from the beginning directly to the end Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 18
HAPP – Case 1 l Constraint Functions l Vehicle v can not go from an activity u to the beginning l If u is an activity, vehicle v can not be finished after u l If v is finished, it can not go to another activity Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 19
HAPP – Case 1 l Summary: § § § Disutility Functions handling trip restrictions Functions handling time restrictions Functions handling demand restrictions Functions handling overall cost and time Functions handling start and stop positions Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 20
HAPP – Case 1 l Example l l l l l 2 Person / Vehicle S = [8, 1, 2] Durations [ai, bi] = [8, 8. 5; 10, 20; 12, 13] [an+i, bn+i] = [17, 19; 10, 21; 12, 21] [a 0, b 0] = [6, 20] [a 2 n+1, b 2 n+1] = [6, 21] Bc = 8 Bt = 3. 5 Ds = 4 Time & Cost Matrixes from activity to activity Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 21
HAPP – Case 1 l Example l Disutility function l Minimize the cost + delay + extent of the travel day Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 22
HAPP – Case 1 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 23
HAPP – Case 2 l Case 1 l l l Unrealistic Only certain people can perform some activities Case 2 l l l Each member of the household has exclusive unrestricted use of a vehicle Some activities can be completed by any member of the household The remaining activities can be completed by a subset of the household members Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 24
HAPP – Case 2 l Constraint Functions l l l This new constraint can be added with new vectors of what activities can not be performed by individual members Thus only one constraint function need be added If a member of the household can not perform w then there is no trip to w Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 25
HAPP – Case 2 l Example l Same as Example 1 with the following added Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 26
HAPP – Case 2 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 27
HAPP – Case 3 l Case 2 l l Better, but still unrealistic Some members of the household should be allowed to stay home. The disutility function should reflect the cost of leaving the house Case 3 l l Each member of the household has exclusive unrestricted use of a vehicle Some activities can be completed by any member of the household The remaining activities can be completed by a subset of the household members A member of the household may perform no activities Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 28
HAPP – Case 3 l Constraint Functions l Recall: l Vehicle v will travel to at least 1 activity l Vehicle v will return home l Replace with: Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 29
HAPP – Case 3 l Example l l l Same as Example 1 with the following added Ω = {null} [ai, bi] = [8, 8. 5; 6, 20; 12, 22] Add 1 more term to the disutility function Where K is the cost associated with leaving the house, in this case 100 was used Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 30
HAPP – Case 3 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 31
HAPP – Case 4 l Case 3 l l Not everyone has unrestricted access to a vehicle Case 4 l l l l Each member of the household has access to a stable of vehicles Some vehicles can be used by any member of te household The remaining vehicles may be used by a subset of members Some activities can be completed by any member of the household The remaining activities can be completed by a subset of the household members Some members of the household may perform no activities Some vehicles may not be used Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 32
HAPP – Case 4 l Decoupling Household Members and Vehicles l Simply need to add household members and their constraints l Household Members Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 33
HAPP – Case 4 l Constraint Functions Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 34
HAPP – Case 4 l Constraint Functions Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 35
HAPP – Case 4 l Constraint Functions l If a household member goes from activity u to activity w then they take a vehicle l A household member must leave home in a vehicle Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 36
HAPP – Case 4 l Example a l Same as Example 3 with the following added Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 37
HAPP – Case 4 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 38
HAPP – Case 4 l Example b l Same as example 4 a with the following changed restrictions on who can perform activities and what vehicles can perform what activities Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 39
HAPP – Case 4 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 40
HAPP – Case 5 l l l l l General HAPP Case Add Ridesharing Each member of the household has access to a stable of vehicles Some vehicles can be used by any member of te household The remaining vehicles may be used by a subset of members Some activities can be completed by any member of the household The remaining activities can be completed by a subset of the household members Some members of the household may perform no activities Some vehicles may not be used Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 41
HAPP – Case 5 l Adding Ridesharing l l l Ridesharing significantly changes the problem The basic formulation (constraints) no longer applies However, the structure remains the same and similar constraint functions can be used All vehicles now must have passenger seats Need to include picking up passengers (discretionary) and dropping off passengers (mandatory) Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 42
HAPP – Case 5 l New Terms Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 43
HAPP – Case 5 l Definitions of Terms Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 44
HAPP – Case 5 l Categories of Constraint Functions l l l Vehicle Temporal Household Member Temporal Spatial Connectivity Constraints on Vehicles Spatial Connectivity Constraints on Household Members Capacity, Budget and Participation Constraints Vehicle and Household Member Coupling Constraints Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 45
HAPP – Case 5 l Vehicle Temporal Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 46
HAPP – Case 5 l Household Member Temporal Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 47
HAPP – Case 5 l Spatial Connectivity Constraints on Vehicles l Activities are performed by either the driver or a passenger l Drivers can perform passenger service activities l Passenger activities are performed on a passenger serve trip Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 48
HAPP – Case 5 l Spatial Connectivity Constraints on Vehicles l Passengers may not perform passenger serve activities Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 49
HAPP – Case 5 l Spatial Connectivity Constraints on Vehicles Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 50
HAPP – Case 5 l Spatial Connectivity Constraints on Household Members Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 51
HAPP – Case 5 l Capacity, Budget and Participation Constraints Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 52
HAPP – Case 5 l Vehicle and Household Member Coupling Constraints l Only one person can travel to any activity in a particular seat l Drivers and passengers can be transferred at home l The departure time of a household member must coincide with the departure of the vehicle they are in Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 53
HAPP – Case 5 l Example l l Same as example 4 b with an increase in duration of activity 2 to allow for a viable ridesharing window Capacity of vehicles is sufficient Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 54
HAPP – Case 5 Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 55
HAPP l Runtime l l l Cases 1 – 4 were run using a commercially available software program GAMS ZOOM. Case 5 was solved using GAMS ZOOM on the nonridesharing problem (Case 4) and then that solution was used to generate viable ridesharing options. These options were then optimized temporally. The best of these was then selected. Case 5 example took 3. 5 minutes on a 50 Mhz machine. Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 56
Conclusion l Utility Maximization l l l It is assumed that activities are chosen and scheduled base on a principle utility maximization HAPP provides a mathematical framework similar to the well studied PDPTW problem. The disutility function can be customized to fit specific needs and will allow for different solutions This framework may contain redundancy and/or hidden inconsistency that may need to be worked out This paper is an initial attempt to provide direction for further research Jeremiah Jilk University of California, Irvine ICS 280, Spring 2004 57
- Slides: 57