Robust Execution of Probabilistic Temporal Plans
Authors: Kyle Lund, Sam Dietrich, Scott Chow, James Boerkoel
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show empirically that our dynamic approach outperforms all known approaches in terms of execution success rate. |
| Researcher Affiliation | Academia | Kyle Lund, Sam Dietrich, Scott Chow, and James C. Boerkoel Jr. Human Experience & Agent Teamwork Laboratory (http://cs.hmc.edu/HEAT/) Harvey Mudd College, Claremont, CA {klund, sdietrich, schow, boerkoel}@hmc.edu |
| Pseudocode | Yes | Algorithm 1: Static Robust Execution Algorithm |
| Open Source Code | No | Simulation code and problem instances available upon request. |
| Open Datasets | No | The paper uses a 'random robot navigation problem generator' and states 'problem instances available upon request', which does not constitute concrete access to a publicly available or open dataset. |
| Dataset Splits | No | The paper describes simulating execution and sampling values, but does not provide specific details on train/validation/test dataset splits (e.g., percentages, sample counts, or predefined splits) for reproducibility. |
| Hardware Specification | Yes | We ran simulations on a Linux machine with 96 Xeon E7540 cores. |
| Software Dependencies | No | The paper states 'We implemented our approaches in Python using the Pu LP linear programming library.' but does not provide specific version numbers for Python or PuLP, or any other software dependencies. |
| Experiment Setup | Yes | We adapt the random robot navigation problem generator of Brooks et al. (2015) to generate random PSTNs with varying numbers of timepoints and constraint characteristics. Structurally, each PSTN is composed of several agent subproblems... The PDFs associated with contingent edges were all normal distributions, with means of less than 10 seconds. The standard deviations (σ) of these distributions were systematically varied using values between 1 and 5... We also varied the interagent constraint density... Finally, the degree of synchronization represents the tightness of the bounds on interagent constraints, which were set to [0, n σ] with n = 1, 2, 4 (n = 1 by default). We find the minimal α (i.e. maximum guaranteed robustness) such that strong controllability can still be established. We minimize α by running a binary search over the rang [0, 1] for a given resolution r, evaluating the LP at every point... we assign risk to its best setting by performing a binary search for the minimal risk budget that still permits a strongly controllable STNU. |