AutonomousRobotsLab

Validated extensively on rotary-wing and ground robots such as multirotors and legged robots (while several of its modules are also tested on fixed-wing aircraft and underwater ROVs), the stack has demonstrated resilient autonomy in GPS-denied and challenging field conditions.

Navigation & Multi-layered Safety: combining map-based collision avoidance and reactive navigation — including (a) Neural SDF-based NMPC (ensuring collision-free motion even in unknown or perceptually degraded spaces), (b) Exteroceptive Deep RL, and (c) CBF-based safety filters.

We’re excited to open-source the *Unified Autonomy Stack* - a step toward a common blueprint for autonomy across robot configurations in the air, on land (and soon at sea). 📦 Repository: github.com/ntnu-arl/unifi… 📘 Documentation: ntnu-arl.github.io/unified_autono… #Robotics #autonomy

UniPilot: Enabling GPS-Denied Autonomy across Embodiments youtu.be/6pwI4-YCV8M?si… via @YouTube

@timjcarden .ee

Starting with light and finishing in pitch darkness - all while mapping most of NTNU's main campus. Long-term collaborative exploration through a marsupial legged-aerial robot team youtu.be/eiRNcaPM5L8?si…

The paper for our open-sourced "Aerial Gym Simulator" is now accepted at IEEE Robotics & Automation Letters and available on Arxiv. * Video: youtu.be/VBVBQNtHZoc * Paper: arxiv.org/abs/2503.01471 * Code: github.com/ntnu-arl/aeria…

🚀 New Paper Alert! 🌊🤖 We are excited to share our latest research accepted for presentation at ICRA2025 on DeepVL Dynamics and Inertial Measurements-based Deep Velocity Learning for Underwater Odometry! ntnu-arl.github.io/deepvl-deep-ve… youtu.be/ctcbrNu_N78

Drone-level vision-driven autonomous exploration, now underwater. Results from field deployment in an unfinished submarine base. Significant step towards a *Unified Resilient Autonomy* architecture. Details soon! youtu.be/CDn_PnH3f5o?si… #robotics #research

Our work on "Morphy: A Compliant and Morphologically Aware Flying Robot" is out! Have a look at the paper doi.org/10.1002/aisy.2… and enjoy the video: youtu.be/C6l7Vklbc9k This work was supported by RCN (grant NO-321435) and @EU_HaDEA ("SPEAR" grant EC 101119774). #robotics

Newer capabilities to query face & vertex indices and surface normal information are added to the rendering framework alongside the capability to query user-defined vertex-level annotations from the environment. Scripts to deploy trained policies on real robots are provided.

We provide scripts for training: * State-based RL policies for position setpoint tracking * Vision based RL-policies for cluttered navigation * Joint-aware policies for soft drones minimizing joint oscillation * Joint & motor control policies for reconfigurable systems

We are happy to release a major update for the "Aerial Gym Simulator". * Code:  github.com/ntnu-arl/aeria… * Video: youtu.be/ROQ8iK_l1jM The work was part of the @spear_robotics project funded by @EU_HaDEA and @AFOSR. #robotics #simulation #autonomy #deeplearning

If you are at #CoRL 2024 this week, come have a look at our work on "In-Flight Attitude Control of a Quadruped using Deep Reinforcement Learning": Open-hardware (low-cost), open-software quadruped with RL for in-flight attitude stabilization.

If you are interested in vision-driven underwater autonomy and looking for a VIO system that doesn't require calibration underwater and instead estimates online the refractive index check our new open-source ReAqROVIO: Refractive Aquatic ROVIO github.com/ntnu-arl/reaqr… #rotics