NAC-MIEL: Source Code for Neural Variational Adaptive Control in Physical Human–Robot Interaction

Abstract

NAC-MIEL: Source Code for Neural Variational Adaptive Control in Physical Human–Robot Interaction This repository archives the original research source code associated with the manuscript "MIEL Framework for Neural Variational Adaptive Control" (Manuscript ID: PONE-D-26-14196), submitted to PLOS ONE. The study proposes NAC-MIEL, a neural variational adaptive control framework for physical human–robot interaction. The framework integrates latent consistency models (LCM) for low-latency action generation, uncertainty-gated fusion (UGF) for multimodal perception, dynamic spatio-temporal hypergraph reasoning for collaborative intention understanding, and variational user modeling with adaptive control barrier functions (CBF) for safety-aware robot control. The research evaluates the framework using the Robomimic manipulation benchmark, the NTU RGB+D 120 interaction subset, and human–robot collaborative assembly experiments involving a Franka Panda robotic platform and NVIDIA Jetson Orin Nano edge computing. This software archive accompanies the manuscript and provides the research source code made available by the authors to support methodological transparency, reproducibility, and further academic investigation. Please