Learning Continuous Environment Fields via Implicit Functions

Authors: Xueting Li, Shalini De Mello, Xiaolong Wang, Ming-Hsuan Yang, Jan Kautz, Sifei Liu

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 EXPERIMENTS", "We evaluate the proposed environment field on agent navigation in 2D mazes (Section 5.2) and compare with the VIN (Tamar et al., 2016). For dynamic human motion modeling in 3D scenes (Section 5.3), we compare with the HMP (Cao et al., 2020) and continuous samplingbased methods RRT (Bry & Roy, 2011) and PRM (Kavraki et al., 1996).
Researcher Affiliation Collaboration Xueting Li1, Shalini De Mello2, Xiaolong Wang3, Ming-Hsuan Yang1, Jan Kautz2, and Sifei Liu2 1UC Merced 2NVIDIA 3UC San Diego
Pseudocode Yes Algorithm 1 Multiple humans navigation.
Open Source Code No We plan to release the code.
Open Datasets Yes Our method is evaluated on publicly avilable dataset (Chevalier Boisvert et al., 2018; Tamar et al., 2016; Hassan et al., 2019).
Dataset Splits Yes We use the same train and test split as VIN, and learn different implicit functions for mazes of different sizes." and "We use the same train and test trajectories as the HMP method
Hardware Specification No The paper does not explicitly describe the specific hardware used for running its experiments, such as GPU models, CPU types, or memory specifications.
Software Dependencies No The paper mentions 'Py Torch framework (Paszke et al., 2019)' but does not specify its version number or any other software dependencies with specific version numbers.
Experiment Setup Yes We train the conditional VAE described in Section 4.2 and all implicit functions using the Adam solver (Kingma & Ba, 2014) with a learning rate of 5e-5. To balance the KL-divergence and the reconstruction objectives in the VAE, we adopt the Cyclical Annealing Schedule introduced in (Fu et al., 2019).