Semi-Black Box: Rapid Development of Planning Based Solutions
Authors: Michael Katz, Dany Moshkovich, Erez Karpas
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our empirical evaluation shows that these heuristics allow the planner to scale up significantly better than the traditional black box approach. |
| Researcher Affiliation | Collaboration | Michael Katz IBM T.J. Watson Research Center Yorktown Heights, NY, USA michael.katz1@ibm.com Dany Moshkovich IBM Watson Health Haifa, Israel mdany@il.ibm.com Erez Karpas Technion Haifa, Israel karpase@technion.ac.il |
| Pseudocode | No | The paper provides code examples in Java (e.g., Figure 1, 3, 4, 5) but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states 'The PDDL domain and problem instances are available upon request.' but does not provide concrete access to the source code for the described methodology. |
| Open Datasets | No | The paper mentions '25 generated problems of an increasing size of the commuter pooling domain' and 'an evaluation of the Evolution domain' and states 'The PDDL domain and problem instances are available upon request.' but does not provide concrete access information for a publicly available dataset. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, or detailed splitting methodology). |
| Hardware Specification | Yes | We used a 2GB memory bound and 30 minutes time bound on a single core of an Intel(R) Core(TM) i7 2.5 GHz machine. |
| Software Dependencies | No | The paper mentions implementing the approach 'in Java' and using 'optic', 'Fast Downward planning framework', 'LAMA planner', and 'FF heuristic' but does not provide specific version numbers for these software dependencies. |
| Experiment Setup | Yes | Our approach (SBB in Table 1) performs an iterative search with found solution cost passed as an upper bound to the next iteration, similarly to the LAMA planner (Richter and Westphal 2010). We start with a greedy best first search, and then weighted A with decreasing weights 5, 3, 2, and 1, continuing with weight 1 until no solution is found. |