Bisimulation Metric for Model Predictive Control

Authors: Yutaka Shimizu, Masayoshi Tomizuka

ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate BS-MPC on both continuous control and image-based tasks from the Deep Mind Control Suite, demonstrating superior performance and robustness compared to state-of-the-art baseline methods.
Researcher Affiliation Academia Yutaka Shimizu & Masayoshi Tomizuka Mechanical Engineering University of California, Berkeley Berkeley, CA 94720, USA EMAIL
Pseudocode Yes C BS-MPC TRAINING ALGORITHM FLOW. The training algorithm flow is described in Algorithm. 1. Algorithm 1 BS-MPC (Model training) Algorithm 2 TD-MPC (Model training)
Open Source Code Yes We publicly release the value of episode return at each time step and code for training BS-MPC agents.
Open Datasets Yes We evaluate BS-MPC across various continuous control tasks using the Deep Mind Control Suite (DM Control (Tassa et al., 2018)). Finally, we evaluate the robustness of the proposed method in the presence of distracting information... Following (Zhang et al., 2018; 2021), driving videos from the Kinetics dataset (Kay et al., 2017) are used as background for the original images.
Dataset Splits No The paper does not provide specific training/validation/test dataset splits for reproducibility in the traditional supervised learning sense. For evaluation, it mentions: "At each evaluation step, the episode return is computed over 10 episodes. The results are averaged over 3 seeds," but this refers to evaluation methodology rather than a static dataset partitioning for training.
Hardware Specification Yes Table. 1 shows the computational time of BS-MPC and TD-MPC. We use RTX-4090 for our experiments.
Software Dependencies No The paper mentions several algorithms and frameworks (e.g., MPPI, SAC, Dreamer-v3, TD-MPC), but it does not specify the version numbers of any software libraries, programming languages, or specific implementations used for its own methodology (BS-MPC).
Experiment Setup Yes D IMPLEMENTATION DETAILS. Here we give details about the hyper-parameters and model architectures. Table 2: Hyperparameters used for TD-MPC and BS-MPC in the experiment. Table 3: Bisimulation metric parameter used in the experiment. Table 4: Model Architecture used in the experiment.