Drone Tech I Like
Deep Reinforcement Learning applied to Robotics
Back in July 2023, I was accepted into the International Computer Vision Summer School (ICVSS 2023), which took place in Sicily, Italy. There were 440 applicants, and only around 120 were selected to attend. For a week, I was locked down in a Sicilian beach resort with some of the brightest Ph.D. students and postdocs in the field of Artificial Intelligence. At 27 years old, I was well above the average age of the attendees. At the airport, I ran into a familiar face—Lorenzo, an Italian researcher who had previously worked in the Motion Capture lab I helped establish in Abu Dhabi. Over the past few years, he had been working with PULP and the Crazyflie nano drone platform alongside another Lorenzo, a colleague of mine in Abu Dhabi. Incidentally, this other Lorenzo went on to win a prestigious nano drone autonomous navigation award—though I can’t recall the exact name of the award right now.
The conference kicked off on Monday with a stellar keynote by Andrej Karpathy, the Head of AI at OpenAI, who, to my surprise, was something of a rockstar in the AI world. In his talk, he broke down and explained, step by step, how ChatGPT was built—an incredible thing to witness.
Throughout the week, various world experts gave keynote sessions, though, to be honest, I didn’t understand much of them. My background and work experience weren’t directly in AI research, so some of the technical discussions went over my head.
But then, on Wednesday morning, everything changed.
That day was dedicated to AI and Computer Vision applied to robotics systems—finally, something I could relate to and understand! This was next-level brain food for me. I was fascinated by the sensorimotor approach to robotics and how top researchers in the field trained control policies in simulation before deploying them to real-world robots.
One of the speakers from Apple demonstrated how, after countless hours of testing and data collection (something I know very well), their robot could hike through the Swiss mountains autonomously—and at nearly the same pace as a human. This research was even featured in Science Robotics.
Next topics
- sensorimotor approach of Berkley professor
- first sneak peak in the world of Davide Scaramuzza's AI FPV drone that could beat the world champion
- Having wine every day for free during dinner
- Exploring Ragusa Ibla and beautiful Sicily
- Making friends and going out in Catania
- Hiking in the Etna Volcano, staying for a night in Taormina and the theater
- Driving around the Volcano to reach Bronte, the land of the sicilian green gold and the best Pistacchio
- Missing europe and the street life