Using Machine Learning for Decreasing State Uncertainty in Planning
Authors: Senka Krivic, Michael Cashmore, Daniele Magazzeni, Sandor Szedmak, Justus Piater
JAIR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results also demonstrate that using our active learning process for identifying information to be sensed leads to gathering information that improves the prediction process. (...) In Section 5 we present the experimental results. |
| Researcher Affiliation | Academia | Senka Krivic EMAIL King s College London (...) Michael Cashmore EMAIL University of Strathclyde (...) Daniele Magazzeni EMAIL King s College London (...) Sandor Szedmak EMAIL Aalto University (...) Justus Piater EMAIL University of Innsbruck (...) |
| Pseudocode | No | The paper describes procedures in numbered steps within paragraphs (e.g., Section 3.1, Section 4.1) but does not present any structured pseudocode or algorithm blocks with typical code-like formatting (e.g., indentation, keywords like 'if', 'for', 'return'). |
| Open Source Code | Yes | The system can be easily used with other domains as well (Krivic, Cashmore, Magazzeni, Ridder, Szedmak, & Piater, 2018). Krivic, S., Cashmore, M., Magazzeni, D., Ridder, B., Szedmak, S., & Piater, J. (2018). State Predictions System together with Domains and Test Scripts. https://github.com/Senka2112/State Predictions. |
| Open Datasets | Yes | The domains TPP, blocksworld, openstacks, and satellite were taken from the International planning Competition (Pommerening, Torralba, & Balyo, 2018). |
| Dataset Splits | Yes | The percentages of knowledge used as known data of the state in tests are 0.5%, 1%, 2%, 3%, 5%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, and 80%. (...) To examine the reproducibility of prediction problems, we randomly generated 10 states for each combination of the percentage of knowledge and problem size. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU models, GPU models, memory, etc.) used for running the experiments. |
| Software Dependencies | No | We integrated the prediction process into the planning and execution framework ROSPlan (Cashmore et al., 2015) in the Robot Operating System (ROS) (Quigley et al., 2009). The contingent Closed-Loop Greedy Planner (CLG) (Albore & Geffner, 2009) and POPF (Coles et al., 2010) are used to solve the resulting planning problems. (...) The resulting plans were validated against the ground truth using VAL (Howey, Long, & Fox, 2004). (No specific version numbers are provided for any of these software components.) |
| Experiment Setup | Yes | The confidence value for completing the graph with predictions was set to ct = 0 in all experiments. (...) A time limit of 1800 [s] was given to planning (including any prediction). (...) The value of ϵ = 0.05 is set for ϵ-greedy exploration for all domains. |