1 / 10

Problem Description & Assumption

Problem Description & Assumption. Metric model in Sapa planner: f(p) = w * time(p) + (1-w) * cost(p). Assuming that the trade-off value w is given. In reality, it's hard to extract the user preference model exactly It's easier for them to say:

bterrell
Download Presentation

Problem Description & Assumption

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Problem Description & Assumption • Metric model in Sapa planner: • f(p) = w * time(p) + (1-w) * cost(p). • Assuming that the trade-off value w is given. • In reality, it's hard to extract the user preference model exactly • It's easier for them to say: • “Yes, I'm interested in time and cost of the travel plan. Let me see some plans, I will choose” (no trade-off preference)‏ • “Hmm, I don't have much money.” (prefer plans with cheap cost)‏ • “I want plans not too expensive, and not too slow as well”. (compromised time and cost)‏

  2. Problem Description & Assumption • Our work's assumption: • User concerns about time, cost of executing a plan. • User preference model is convex combination between two objectives • f(p) = w * time(p) + (1-w) * cost(p)‏ • One of the most frequently used model. • The trade-off between time, cost is unknown. • But the distribution of w is given. • Problem: • Find a set of plans such that the user can select one satisfying their hidden trade-off.

  3. Solution Approach • Integrated Convex Preference (ICP) measure to evaluate the quality of a set of plan X. • If ICP(X1) < ICP(X2) then X1 is better than X2 in supporting the user: • Given a trade-off value w of the user, • X1 is more likely contains better quality plan than X2. • Extend the A* search procedure of Sapa: • Consider 2 objectives: Multi-objective A*. • Put heuristic on top of Multi-objective A*: • To select the most promising node

  4. Multi-objective A* with ICP measure

  5. How well the set can be?

  6. How to test the approach? • Sapa with sampling: • Generate set of N w values, based on the distribution: {w1, ..., wN} • For each w_i, invoke Sapa to find a plan p(w_i). • P={p(w1), ..., p(wN)}. Time: t1. • Our approach: Q = {q1, ..., qM}. Time: t2. • Comparison: • Simulating T transactions. • For each transaction t, generate w based on distribution. • Compare 2 optimal plans w.r.t w in P and Q. • Compare t1, t2.

  7. Future works(though current work is on-going ;-)‏ • Qualitative preference model • On trajectory / behavior of the plans. For instance: prefer United Airline to American Airline; would like to visit some places during the trip, ... • Our work as Over-subscription planning: • Utility is fixed: all goals are achievable. • Cost is more general. • Natural extension: utility as an objective to maximize.

  8. Thank you!Q &A

More Related