Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Dynamic Controllability of Controllable Conditional Temporal Problems with Uncertainty
Authors: Jing Cui, Patrik Haslum
JAIR 2019 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We illustrate the benefit of fully dynamic strategies for CCTPUs by comparing implementations of DC checking methods for the CCTPU with and without dynamic discrete choices. The DC checking method for the CCTPU which does not make assignments to discrete variables dynamically only considers scheduling time points dynamically. Our implementation of this method checks each leaf node (full set of assignments) in turn and considers the CCTPU dynamically controllable if it finds one leaf that induces a dynamically controllable STPU. The result is shown in Figure 22. The tests are grouped by DC checking results in the chart: infeasible means the problem is not dynamically controllable; feasible with fixed option means the algorithm found a viable strategy with a static assignment of the discrete variables; and feasible with dynamic options means it found a strategy that both assigns discrete variables and schedules time points dynamically. |
| Researcher Affiliation | Academia | Jing Cui EMAIL Patrik Haslum EMAIL Research School of Computer Science, Australian National University & Decision Sciences Programs, DATA61, CSIRO 108 North Rd, Acton ACT 2601, Australia |
| Pseudocode | Yes | The approach to extract a complete set of conflicts is summarised in Algorithms 1, 2 and 3. Algorithm 1: Extracting the DC envelope of an STPU. Algorithm 2: Modified backpropagation process that can extract all conflicts. Algorithm 3: The DFS process that do reductions through positive links. Algorithm: Tree Search(Node, A, S). Algorithm: Execute CCTPU |
| Open Source Code | No | The paper mentions using "the benchmark generator by Yu (2016)" and cites "Yu, P. (2016). BCDR Test Generator. https://github.com/yu-peng/BCDRTestGenerator." This refers to a third-party tool used for generating test cases, not the authors' own implementation of their described methodology for open-source release. |
| Open Datasets | No | We use the benchmark generator by Yu (2016), based on Zipcar problems (Yu & Williams, 2013; Yu et al., 2014). Its application background is a car-sharing network. Each test case consists of missions with temporal requirements, each mission has a sequence of activities, and each activity can be done by choosing one option... We generated 16000 test cases, with 1 8 discrete variables and 1 10 options for each variable. The paper describes generating its own test cases using a benchmark generator but does not provide a specific link, DOI, or repository for the 16000 generated test cases used in the experiments. |
| Dataset Splits | No | We generated 16000 test cases, with 1 8 discrete variables and 1 10 options for each variable. The tests are grouped by DC checking results in the chart: infeasible means the problem is not dynamically controllable; feasible with fixed option means the algorithm found a viable strategy with a static assignment of the discrete variables; and feasible with dynamic options means it found a strategy that both assigns discrete variables and schedules time points dynamically. The paper mentions |
| Hardware Specification | Yes | Zavatteri (2017) solved an example with two decisions, 4 contingent links and 11 nodes by implementing the TGA in UPPAAL-TIGA, which took about one minute (running on a virtual machine with Intel i7 2.8GHz CPU and 5G RAM). Solving a problem of the same size with the DC checking algorithm in this paper needs less than a second (running on a desktop with Intel i5 3.2GHz CPU and 4G RAM). |
| Software Dependencies | No | The paper does not provide specific version numbers for software libraries, programming languages, or environments used for their own implementation. It mentions UPPAAL-TIGA in a comparative discussion of related work, but not as a dependency for their core methodology. |
| Experiment Setup | Yes | All temporal links are randomly generated except for the requirement on the overall duration of the missions, which randomly deviates by 20% from the estimated bounds of the sequence of activities... In this experiment, we set the lower bounds of the contingent links to their original lower bounds and upper bounds are the sum of their lower bounds and the maximum deviation. |