SkyJEPA model enables zero-shot sim-to-real quadrotor control

By PulseAugur Editorial · [1 sources] · 2026-06-22 15:00

Researchers have developed SkyJEPA, a novel approach to learning long-horizon world models for quadrotor control. This JEPA-style model operates in latent space and incorporates a physics-inspired prober to map frozen latents to interpretable states, enabling physically grounded predictions. The system is designed for real-time control on embedded hardware and utilizes automated dataset generation to reduce reliance on real-world data collection. Experiments show accurate prediction, robust zero-shot sim-to-real transfer, and strong generalization capabilities. AI

IMPACT This research advances the capability of quadrotors to perform complex, long-horizon tasks by improving predictive modeling and sim-to-real transfer.

RANK_REASON The cluster contains a research paper detailing a new model and its experimental validation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

SkyJEPA model enables zero-shot sim-to-real quadrotor control

COVERAGE [1]

arXiv cs.LG TIER_1 English(EN) · Giuseppe Loianno · 2026-06-22 15:00

SkyJEPA: Learning Long-Horizon World Models for Zero-Shot Sim-to-Real Control of Quadrotors

Accurate dynamics models are critical for informed decision-making in robotic systems, particularly for agile aerial vehicles operating under uncertainty. Neural network dynamics models are attractive for capturing complex nonlinear effects, but existing predictive approaches str…

COVERAGE [1]

SkyJEPA: Learning Long-Horizon World Models for Zero-Shot Sim-to-Real Control of Quadrotors

RELATED ENTITIES

RELATED TOPICS