Adaptive Information Belief Space Planning

Authors: Moran Barenboim, Vadim Indelman

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments The goal of the experiment section is to evaluate the influence of the abstraction mechanism on the planning performance. We examined both the time difference and the total return. All algorithms use a particle filter for inference, the choice of the particle filter variant is independent of our contribution. All experiments were performed on the common 2D Light Dark benchmark, where both the state and observation spaces are continuous; see an illustration in Figure 4.
Researcher Affiliation Academia Moran Barenboim1 and Vadim Indelman2 1Technion Autonomous Systems Program 2Department of Aerospace Engineering Technion Israel Institute of Technology, Haifa 32000, Israel moranbar@campus.technion.ac.il, vadim.indelman@technion.ac.il
Pseudocode Yes Algorithm 1 AI-FSSS Procedure: SIMULATE(b,d) (...) Algorithm 2 SOLVE Procedure: SOLVE (...) Algorithm 3 REFINE Procedure: REFINE(b,ba,d)
Open Source Code Yes For the appendix and the code please visit https://github.com/ moranbar/Adaptive-Information-BSP
Open Datasets No All experiments were performed on the common 2D Light Dark benchmark, where both the state and observation spaces are continuous; see an illustration in Figure 4. The paper mentions the benchmark but does not provide a specific link, DOI, or formal citation (with author and year) for accessing its dataset.
Dataset Splits No The paper mentions using a 'particle filter' and performing 'simulations' and 'full trajectories' but does not provide specific train/validation/test dataset splits (exact percentages, sample counts, or citations to predefined splits).
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions 'particle filter' and 'MCTS' but does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes In the first experiment, we fixed the number of state particles to n = 40, and examined the influence of different branching factors over the observation space, see Figure 3. The algorithms were limited to 20,000 iterations before performing an action. (...) Here, the number of observations was fixed to M = 4. (...) All algorithms had 1 second limitation for planning before each interaction with the environment. (...) Each algorithm performed 1,000 full trajectories in the environment, each contained 25 steps.