Continual Planning in Golog
Authors: Till Hofmann, Tim Niemueller, Jens Claßen, Gerhard Lakemeyer
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on autonomous mobile robots show that the approach supports expressive behavior specification combined with efficient sub-plan generation to handle dynamic environments and incomplete knowledge in a unified way. In this section, we evaluate our approach to continual planning in GOLOG with two applications. We present an indepth evaluation of our approach in the household domain and compare it to the GKI CP. |
| Researcher Affiliation | Academia | Till Hofmann, Tim Niemueller, Jens Claßen, and Gerhard Lakemeyer Knowledge-Based Systems Group RWTH Aachen University, Germany {hofmann, niemueller, classen, lakemeyer}@kbsg.rwth-aachen.de |
| Pseudocode | Yes | while some(z, zone, explorable(z)) do plan(some(z,zone,and(explorable(z),explored(z)))) end While |
| Open Source Code | Yes | Our implementation is available on https://www.fawkesrobotics.org/p/golog-cp. |
| Open Datasets | No | We evaluate our approach using the clean-up task from the household domain and compare it to the GKI CP. A comparison to MAPL was not possible because our domain relies on conditional effects, which are not supported by MAPL. To compare both continual planners, we run both systems on the same problem. The paper describes the problem settings and refers to "the Tidy Up-Robot project (Dornhege and Hertle 2013)" but does not provide a specific link or citation to a publicly accessible dataset used for the experiments. |
| Dataset Splits | No | The paper describes experimental tasks and conditions but does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts). |
| Hardware Specification | Yes | All benchmarks were done on an Intel Core i7-3770 at 3.40 GHz with 4 cores and hyper-threading, resulting in 8 parallel threads. |
| Software Dependencies | No | Our GOLOG interpreter is based on the Prolog implementations of INDIGOLOG and READYLOG. We use ECLi PSe as Prolog interpreter which is integrated into the robot software framework Fawkes (Niemueller et al. 2010). The Lua-based Behavior Engine (BE) (Niemueller, Ferrein, and Lakemeyer 2010) provides the primitive actions. The paper names software components like ECLi PSe, Fawkes, and Lua-based Behavior Engine, but does not provide specific version numbers for them. |
| Experiment Setup | Yes | Our evaluation task is a clean-up task from the household domain and modifies the setting of the Tidy Up-Robot project (Dornhege and Hertle 2013): The robot is supposed to clean up all cups located on a table by putting dirty cups into the dish washer and clean cups on the shelf. Initially, the robot knows all locations, how to align to all locations and all the spots in the dishwasher. The robot is able to sense the state of a cup if it is holding the cup with the explicit is cup clean sensing action. It also relies on passive sensing... The goal is always the same... The two tasks only differ in the initial knowledge: Task 1 (The robot initially knows neither the cup positions nor whether a cup is clean. We vary the number of cups between 1 and 10; we add clean and dirty cups alternately.) and Task 2 (The robot does not know the cup positions, but it knows cup states initially and thus does not need to sense them. As before, every second cup is clean, the number of cups varies between 1 and 10). |