evelyd/paper_delia_2025_iros_physics-informed_trajectory_generation_dataset

<h1 align="center"> Stabilizing Humanoid Robot Trajectory Generation via Physics-Informed Learning and Control-Informed Steering </h1>

<div align="center"> Evelyn D'Elia, Paolo Maria Viceconte, Lorenzo Rapetti, Diego Ferigo, Giulio Romualdi, Giuseppe L'Erario, Raffaello Camoriano, and Daniele Pucci <br> <br> </div>

<br>

<div align="center"> Submitted to the 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). </div>

<section> <h2>📂 Dataset</h2>

<p> The dataset is organized in folders. The layout is explained below. </p> <ul> <li> The <code>mocap/D2</code> folder contains the raw motion capture data, acquired via a human using the Xsens motion tracking suit. Within <code>D2</code> there are 5 subfolders for each data subset: forward, backward, side, diagonal, and mixed walking. each subfolder contains a <code>log</code> file which can be retargeted onto the robot. </li> <li> The <code>retargetedmocap</code> folder contains the text files of the retargeted mocap data, in both original (<code>D2</code>) and mirrored (<code>D2mirrored</code>) form. </li> <li> The <code>IOfeatures</code> folder contains the text files of the extracted features (data and labels) to be used in training. There are 4 subfolders which correspond to the original and mirrored inputs and the original and mirrored outputs. </li> <li> The <code>onnx</code> folder contains the onnx files corresponding to trained models from each PI loss value as well as from the original ADHERENT framework. </li> <li> The <code>reproduceplots</code> folder contains <code>.mat</code> files with all the data needed to reproduce the selected plots. </li> </ul> </section>