Privacy Preserving Pattern Databases Academic Article uri icon

abstract

  • Abstract In a privacy preserving multi-agent planning problem agents have some private information they do not wish to share with each other, but operate to achieve common goals. Based on the Multi-Agent STRIPS (MASTRIPS) formalism, several search-based algorithms were proposed to solve these type of planning problems. To guide the search, these algorithms require a heuristic function that estimates the cost of reaching a goal. Developing such heuristics is challenging because the computation of the heuristic may violate the agents privacy. In this work we show how pattern databases, a general-purpose class of heuristic that can be very effective in many domains, can be implemented in a privacy preserving manner. As a secondary contribution, we present an improvement to the GPPP, a recent effective suboptimal privacy preserving planner and also demonstrate how privacy …

publication date

  • January 1, 2015