Belief Manipulation Through Propositional Announcements

Authors: Aaron Hunter, François Schwarzentruber, Eric Tsang

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We then describe an implemented tool that uses announcement finding to control robot behaviour through belief manipulation.
Researcher Affiliation Academia Aaron Hunter BC Institute of Technology Burnaby, Canada aaron hunter@bcit.ca Francois Schwarzentruber ENS Rennes Bruz, France francois.schwarzentruber@ens-rennes.fr Eric Tsang BC Institute of Technology Burnaby, Canada surplus.et@gmail.com
Pseudocode Yes EXIST ANN (K1, . . . , Kn, ψ1, . . . , ψn) 0. Let m be the size of the underlying input vocabulary of K1, . . . , Kn, ψ1, . . . , ψn 1. Guess d1, . . . , dn {0, 1, . . . , m} 2. Guess valuations v1, . . . , vn 3. For all i, j {1, . . . , n} If d(Kj, vi) < dj, reject. 4. For all i {1, . . . , n} If d(Ki, vi) > di, reject. 5. For all i, j {1, . . . , n} If d(Kj, vi) = dj and vi |= ψj, reject. 6. Accept. and FIND ANN DET (K1, . . . , Kn, P1, . . . , Pn, ψ1, . . . , ψn) 0. Let m = maxi,s Pi(s) 1. For each d = (d1, . . . , dn) with 0 di m: 2. For each n-tuple v = (v1, . . . , vn) of valuations: 3. If Pj(vi) < dj for any i, j, continue. 4. If Pi(vi) > di for any i, continue. 5. If Pj(vi) = dj and vi |= ψj for any i, j, continue. 6. Return φ = form( v) 7. Return no solution.
Open Source Code No Ann B is written in Kotlin4. It is built on the publicly available libraries of Gen B [Hunter and Tsang, 2016]. Briefly, Gen B is a general belief revision solver that is able to calculate the result of any AGM belief revision operation. The paper mentions that their tool Ann B is built on publicly available libraries (Gen B), but it does not provide a link or explicit statement about the source code for Ann B itself being open source or provided.
Open Datasets No The paper describes a simulated environment for robot control and does not refer to the use of any publicly available or open datasets for training, validation, or testing. The example involves a vocabulary `{patrol, checkgate}` and initial belief states, which are internal to the simulation.
Dataset Splits No The paper describes a theoretical framework with a simulated demonstrator and does not specify training, validation, or test dataset splits. The problem is framed as a search for a formula, not a machine learning task with data splits.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments or the simulated robot controller.
Software Dependencies No Ann B is written in Kotlin4. It is built on the publicly available libraries of Gen B [Hunter and Tsang, 2016]. The footnote states: 'Kotlin is a variant of java, available at www.kotlinlang.org.' While Kotlin and Gen B are mentioned, specific version numbers for either are not provided in the paper.
Experiment Setup No The paper describes the functionality and algorithm (FIND ANN DET) of the Ann B tool, including how agents and behaviors are defined. However, it does not provide specific numerical hyperparameters (e.g., learning rates, batch sizes) or system-level training configurations typically found in experimental setups, as it focuses on the logic and algorithmic procedure rather than empirical tuning.