Planning to Avoid Side Effects
Authors: Toryn Q. Klassen, Sheila A. McIlraith, Christian Muise, Jarvis Xu9830-9839
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper we investigate how to avoid side effects in a symbolic planning setting. We study the notion of minimizing side effects in the context of a planning environment where multiple independent agents co-exist. We define (classes of) negative side effects in terms of their effect on the agency of those other agents. Finally, we show how plans which minimize side effects of different types can be computed via compilations to cost-optimizing symbolic planning, and investigate experimentally. |
| Researcher Affiliation | Academia | Toryn Q. Klassen1,2,3, Sheila A. Mc Ilraith1,2,3, Christian Muise4, Jarvis Xu4 1Department of Computer Science, University of Toronto, Toronto, Canada 2Vector Institute for Artificial Intelligence, Toronto, Canada 3Schwartz Reisman Institute for Technology and Society, Toronto, Canada 4School of Computing, Queen s University, Kingston, Canada toryn@cs.toronto.edu, sheila@cs.toronto.edu, christian.muise@queensu.ca, 15gx3@queensu.ca |
| Pseudocode | No | The paper provides formal definitions for compilations (e.g., Definition 18, 21, 22), but these are presented as mathematical specifications of transformations, not in a pseudocode format or labeled as an algorithm block. |
| Open Source Code | Yes | See https://github.com/tqk/side-effects-planner for the code for the experiments, which was written in Python and heavily relies on the Tarski library (https://github.com/aig-upf/tarski). |
| Open Datasets | Yes | We tested on a formalization of our Canadian wildlife domain (from Figure 1), and adaptations of the standard IPC planning domains zenotravel, floortile, and storage (see Appendix B). |
| Dataset Splits | No | The paper does not provide specific training/validation/test dataset split information. It mentions using "standard IPC planning domains" but does not detail how data within these domains was partitioned for training, validation, or testing purposes. |
| Hardware Specification | Yes | The experiments were run on a Linux workstation with a Core i9-9900K CPU (3.60 GHz) and 32GB of RAM. |
| Software Dependencies | Yes | We used the LM-Cut planner (Helmert and Domshlak 2011) for all problems except the goal-preserving plans whose compilations had conditional effects, necessitating LAMA (Richter and Westphal 2010), which we ran to completion instead. Both LM-Cut and LAMA are configurations of Fast Downward (Helmert 2006) (version 20.06 was used). |
| Experiment Setup | Yes | We used the LM-Cut planner (Helmert and Domshlak 2011) for all problems except the goal-preserving plans whose compilations had conditional effects, necessitating LAMA (Richter and Westphal 2010), which we ran to completion instead. Both LM-Cut and LAMA are configurations of Fast Downward (Helmert 2006) (version 20.06 was used). |