Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generalising Planning Environment Redesign

Alberto Pozanco, Ramon Fraga Pereira, Daniel Borrajo

TL;DR

This work generalizes Planning Environment Redesign beyond goal/plan recognition by introducing a metric-agnostic framework, and presents ger, an anytime BFS-based method that prunes the action-removal space using a plan-library generated by a top-quality planner. By supporting a broad class of redesign metrics, including goal/plan transparency and privacy as well as distance-based objectives, ger demonstrates strong empirical performance and scalability. The authors show that a single, unified approach can outperform metric-specific GRD methods on standard tasks and adequately handle novel metrics, enabling flexible, multi-objective planning environment redesign. The results suggest significant practical impact for designing environments that shape agent behavior under diverse objectives, with potential applications in anticipatory planning, counterplanning, and risk management.ger achieves these benefits by efficiently computing a plan-library and using it to guide a BFS over environment modifications, offering soundness, completeness, and (under reasonable assumptions) optimality, while delivering substantial speedups across domains.

Abstract

In Environment Design, one interested party seeks to affect another agent's decisions by applying changes to the environment. Most research on planning environment (re)design assumes the interested party's objective is to facilitate the recognition of goals and plans, and search over the space of environment modifications to find the minimal set of changes that simplify those tasks and optimise a particular metric. This search space is usually intractable, so existing approaches devise metric-dependent pruning techniques for performing search more efficiently. This results in approaches that are not able to generalise across different objectives and/or metrics. In this paper, we argue that the interested party could have objectives and metrics that are not necessarily related to recognising agents' goals or plans. Thus, to generalise the task of Planning Environment Redesign, we develop a general environment redesign approach that is metric-agnostic and leverages recent research on top-quality planning to efficiently redesign planning environments according to any interested party's objective and metric. Experiments over a set of environment redesign benchmarks show that our general approach outperforms existing approaches when using well-known metrics, such as facilitating the recognition of goals, as well as its effectiveness when solving environment redesign tasks that optimise a novel set of different metrics.

Generalising Planning Environment Redesign

TL;DR

This work generalizes Planning Environment Redesign beyond goal/plan recognition by introducing a metric-agnostic framework, and presents ger, an anytime BFS-based method that prunes the action-removal space using a plan-library generated by a top-quality planner. By supporting a broad class of redesign metrics, including goal/plan transparency and privacy as well as distance-based objectives, ger demonstrates strong empirical performance and scalability. The authors show that a single, unified approach can outperform metric-specific GRD methods on standard tasks and adequately handle novel metrics, enabling flexible, multi-objective planning environment redesign. The results suggest significant practical impact for designing environments that shape agent behavior under diverse objectives, with potential applications in anticipatory planning, counterplanning, and risk management.ger achieves these benefits by efficiently computing a plan-library and using it to guide a BFS over environment modifications, offering soundness, completeness, and (under reasonable assumptions) optimality, while delivering substantial speedups across domains.

Abstract

In Environment Design, one interested party seeks to affect another agent's decisions by applying changes to the environment. Most research on planning environment (re)design assumes the interested party's objective is to facilitate the recognition of goals and plans, and search over the space of environment modifications to find the minimal set of changes that simplify those tasks and optimise a particular metric. This search space is usually intractable, so existing approaches devise metric-dependent pruning techniques for performing search more efficiently. This results in approaches that are not able to generalise across different objectives and/or metrics. In this paper, we argue that the interested party could have objectives and metrics that are not necessarily related to recognising agents' goals or plans. Thus, to generalise the task of Planning Environment Redesign, we develop a general environment redesign approach that is metric-agnostic and leverages recent research on top-quality planning to efficiently redesign planning environments according to any interested party's objective and metric. Experiments over a set of environment redesign benchmarks show that our general approach outperforms existing approaches when using well-known metrics, such as facilitating the recognition of goals, as well as its effectiveness when solving environment redesign tasks that optimise a novel set of different metrics.
Paper Structure (16 sections, 1 theorem, 11 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 16 sections, 1 theorem, 11 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Redesign Metrics
  2. Goal Transparency (GT).
  3. Plan Transparency (PT).
  4. Goal Privacy (GP).
  5. Plan Privacy (PP).
  6. Minimise Average Distance (MinAvgD).
  7. Maximise Average Distance (MaxAvgD).
  8. Minimise Maximum Distance (MinMaxD).
  9. Maximise Minimum Distance (MaxMinD).
  10. Compute Plan-Library $\mathbb{P}$ (Line \ref{['alg:bfs_redesign:compute_plan_library']}).
  11. Compute Allowed Modifications (Line \ref{['alg:bfs_redesign:compute_allowed_modifications']}).
  12. Searching Process (Lines \ref{['alg:bfs_redesign:search_begin']}--\ref{['alg:bfs_redesign:search_end']}).
  13. Theoretical Properties.
  14. Benchmarks and Setup.
  15. Execution Time Results and Comparison against GRD.
  16. ...and 1 more sections

Key Result

Theorem 1

Let us assume $C=\{ \textsc{open}=\emptyset \}$, $amod = \mathcal{A}$, and infinite planning time and memory resources. Under those assumptions, ger is sound, complete, and optimal.

Figures (4)

  • Figure 1: grid environment where the agent located at cell $(2,0)$ and has two possible goals: $G_1 = (0,4)$; $G_2 = (4,4)$.
  • Figure 2: Redesigned environments for different metrics by using our approach with a time limit of $15$ minutes and a maximum number of removed actions $|\mathcal{A}_\neg|=4$. Red arrows and lines indicate removed actions $\mathcal{A}_\neg$. Intended goals are depicted in grey.
  • Figure 3: Histograms for the redesign metrics, showing the number of improved/optimised problems using ger.
  • Figure 4: Violin plots for the redesign metrics, showing the improvement from $m_0$ to $m^{+}$ ($|m_0 - m^{+}|$) using ger.

Theorems & Definitions (14)

  • Definition 1
  • Definition 2
  • Definition 3
  • Definition 4
  • Definition 5
  • Definition 6
  • Definition 7
  • Definition 8
  • Definition 9
  • Definition 10
  • ...and 4 more