Faster Stackelberg Planning via Symbolic Search and Information Sharing

Abstract

Stackelberg planning is a recent framework where a leader and a follower each choose a plan in the same planning task, the leader's objective being to maximize plan cost for the follower. This formulation naturally captures security-related (leader=defender, follower=attacker) as well as robustness-related (leader=adversarial event, follower=agent) scenarios. Solving Stackelberg planning tasks requires solving many related planning tasks at the follower level (in the worst case, one for every possible leader plan). Here we introduce new methods to tackle this source of complexity, through sharing information across follower tasks. Our evaluation shows that these methods can significantly reduce both the time needed to solve follower tasks and the number of follower tasks that need to be solved in the first place.

Optimal Mixed Strategies for Cost-Adversarial Planning Games

“…Brafman et al (2009 in-troduce coalition games into multiagent setting. Stackelberg game where a leader executes a plan followed by a plan of the follower were considered in (Speicher et al 2018;Torralba et al 2021). In this paper, we focus on a conceptually much easier setting where the adversary agent directly modifies the action costs.…”

On the Computational Complexity of Stackelberg Planning and Meta-Operator Verification

“…As the number of possible plans is exponential in the worst case, this makes one wonder how the complexity of Stackelberg planning relates to classical planning. Focusing on algorithmic improvements (Speicher et al 2018a;Torralba et al 2021;Sauer et al 2023), existing works have neglected this question so far.…”

Effective Planning in Resource-Competition Problems by Task Decomposition

“…For example, Plan Interdiction Games leverage planning to reason about attackers and defenders in computer networks (Vorobeychik and Pritchard 2020). From the domain-independent standpoint, planning has been used in Stackelberg games for generating follower's plans against the adversarial leader (Speicher et al 2018;Torralba et al 2021), or for counter-planning, where an agent tries to invalidate landmarks required by the adversary, so the adversary cannot achieve its goal (Pozanco et al 2018). Recently, a subclass of normal-form games has been tackled from the domain-independent planning perspective (Rytíř, Chrpa, and Bošanský 2019;Chrpa, Rytíř, and Horčík 2020).…”