Learning-oriented Simulation for Aerial Robots

Organizers

Kostas Alexis (NTNU), Erdal Kayacan (UPB), Stefano Mintchev (ETH Zurich), Karine Miras (VUB), George Nikolakopoulos (LTU)

Presenters

• Stefano Mintchev, ETH Zurich, “Simulating Soft Micro Aerial Vehicles”
• Karine Miras, Vrije Universiteit Amsterdam, “Simulation for Robot Evolution: Requirements & Tools”
• Mihir Kulkarni and Kostas Alexis, Norwegian University of Science and Technology, “Aerial Gym 2.0: Isaac Gym-based Massively Parallelized Simulation for Aerial Robot Learning”
• Marcelo Jacinto, "First Steps with the Pegasus Simulator: An Isaac Sim Framework for Aerial Vehicles”
• Leonard Bauersfeld and Davide Scaramuzza, “Learning Agile Quadrotor Flight: from Simulation to Real-World Deployment”

Abstract

Aerial robot simulation tools have experienced a leap in capacity and capability and currently facilitate robust learning of complex control policies through data. Exploiting progress in high-fidelity physics simulation, improved graphics, and the strengths of GPU-accelerated massively parallelized computing, modern simulators enable robots to collect millions of ‘experiences’ in a short time, thus allowing scalable learning and robust transfer to reality. Within this framework, there is a rising interest in simulators that allow the modeling of arbitrary aerial robot embodiments and simultaneously offer the capacity to generatively create environments either by using neural methods or parametric setting of reconfigurable scenes. The purpose is to automate the process of the computational design of novel aerial robots and their embodied AI solutions through efficient cycles of simulation, learning and adaptation.

Relevant Symposium

SSCI-wide

Overview of Topic

High-fidelity, parallelized and scalable simulation is the key for efficient and robust robot learning. Especially as we aim for policies of increased complexity by a diverse set of robot embodiments, general-purpose simulation environments with the capability to seamlessly generate diverse worlds become essential. Task examples include learning end-to-end position control or autonomous vision-driven navigation for aerial robots. At the core of this technology are major breakthroughs in computing hardware and software, from the progress in GPUs to efficient libraries for high-fidelity physics simulation and graphics modeling and rendering. Accordingly, the proposed tutorial is at the heart of applied computational intelligence and, thus, a perfect fit for SSCI.

Schedule