Data-Driven Revision of Conditional Norms in Multi-Agent Systems
Authors: Davide Dell'Anna, Natasha Alechina, Fabiano Dalpiaz, Mehdi Dastani, Brian Logan
JAIR 2022 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate DDNR using a state-of-the-art, off-the-shelf urban traffic simulator. The results show that DDNR synthesises revised norms that are significantly more accurate than the original norms in distinguishing adequate and inadequate behaviors for the achievement of the system-level objectives. |
| Researcher Affiliation | Academia | Davide Dell Anna EMAIL Delft University of Technology, Delft, The Netherlands Natasha Alechina EMAIL Fabiano Dalpiaz EMAIL Mehdi Dastani EMAIL Utrecht University, Utrecht, The Netherlands Brian Logan EMAIL Utrecht University, Utrecht, The Netherlands University of Aberdeen, Aberdeen, United Kingdom |
| Pseudocode | Yes | Algorithm 1 (More Spec) constructs more specific formulas than the input formula ϕ... Algorithm 2 (Less Spec) constructs less specific formulas than the formula ϕ given as input... Algorithm 3 (Synthesis) takes as input a set of traces Γ... Algorithm 4 get States Algorithm 5 DDNR |
| Open Source Code | Yes | We apply and experimentally evaluate a Java implementation of DDNR, available at (Dell Anna et al., 2022a) |
| Open Datasets | Yes | The source code of our implementation of DDNR and the results and material concerning our experiments are available at (Dell Anna et al., 2022a). |
| Dataset Splits | Yes | In the first experiment (75-25 splitting), we use a standard (machine learning) splitting technique: we take the 100 data sets obtained for RQ1 and we split each of them in a training and a test set, composed of 75% and 25% of the traces in the data set, respectively. |
| Hardware Specification | No | The paper does not provide specific hardware details used for running the experiments. It mentions using a 'SUMO traffic simulator' and 'Java implementation' but no CPU/GPU models or other hardware specifications. |
| Software Dependencies | No | The paper mentions using 'SUMO traffic simulator (Krajzewicz et al., 2012)' and 'a Java implementation' but does not specify exact version numbers for SUMO, Java, or any other libraries or frameworks used in the implementation. |
| Experiment Setup | Yes | DDNR Configuration. ...we set x = 8 in our experiments. As a consequence, build Conj can never generate more than 2^8 = 256 conjunctions from propositions belonging to VP, VD, and VS, and never more than 2^2 = 4 from propositions belonging to VT (because |VT| = 2), for a maximum number of possible conjunctions of 2^10 = 1,024. |