The video presenting our ICAPS18 paper Strong Stubborn Sets for Efficient Goal Recognition Design can be found here
Goal Recognition Design
Offline design as a way to facilitate goal recognition
Goal recognition design(grd) is a new problem , in which we take a domain theory and a set of goals and ask the following questions: to what extent do the actions performed by an agent within the model reveal its objective, and what is the best way to modify a model so that any agent acting in the model reveals its objective as early as possible (click here for the papers published so far and here for the grd code for optimal agents and benchmarks). Click here for our GRD repository that includes all the code, benchmarks and GRD problem generator we have used for my PhD thesis and all GRD papers.
The example above shows a simplified airport terminal, where passengers enter through one entrance and are assumed to walk optimally toward one of the exits on each of the corners. For most of the passengers it is possible to recognize the goal according to their observed behavior. For the passenger marked in red, however, this is not the case. This passenger can advance all the way to the opposite wall before revealing her goal.
As a first stage, Goal Recognition Design finds the Worst Case Distinctiveness (wcd) of a model - which is the maximal number of steps (length of a path) an agent can advance in a system without revealing her goal. In this example the wcd is equal to the distance from the entrance to the opposite wall.
As a second stage, after finding the wcd of a model - the objective is to mnimize it. This is done by dissallowing actions. In the example below, the addition of the fountain enforces the passengers to reveal there goal early in the execution of their advaceent towards one of the gates. In particual the passenger in red reveals her goal to be the right door.
Software and Benchmarks
The code and benchmarks that we used in the Goal Recognition Design paper (ICAPS 14) can be downloaded from here :
The code and benchmarks we used for my PhD thesis and ICAPS 18 and JAIR 19 paper can be found here:
The playGRounD game that demonstrates the GRD concepts. Implemented by Ran Harari
Slides of our ICAPS'19 tutorial on Goal Recognition Design by Sarah Keren and William Yeoh
Presenting Goal Recognition Design in ICAPS 2014 - 24th International Conference on Automated Planning and Scheduling
Presenting Goal Recognition Design for Non Optimal Agents in AAAI 15 - The Twenty-Ninth AAAI Conference on Artificial Intelligence
Presenting Privacy preserving plans in partially observable environments. In Proceedings of the International Joint Conference on Articial Intelligence (IJCAI 2016),