Summer Internships at the Intelligent Robotics Group
Every year the Intelligent Robotics Lab hires summer interns for a multitude of projects, ranging from infrastructure projects to advanced research topics. Two summer interns who recently finished their summer stay in Helsinki in 2023 are Tiia Tikkala and Sergio Hernández, both supervised by Dr. Kevin Sebastian Luck.
Designing Robots and Letting Drones Learn to Fly
Tiia Tikkala, a recently graduated Bachelor’s student of Aerospace Engineering at Delft University of Technology, worked at the Intelligent Robotics Lab for three months in the summer of 2023 in the areas of Robot Design and UAVs. Thanks to her unique experience in rocket design and mechanical engineering, she supported ongoing research efforts in the lab by designing and 3D-printing novel robot designs with CAD software. For this, Tiia could utilise the extensive 3D-printing infrastructure at the Department of Electrical Engineering, especially the Ultimaker S5. In addition, Tiia created a machine-learning framework allowing quadcopters to learn to fly autonomously from data. To achieve this, Tiia collected positional data of a Crazyflie drone from the Lab’s Optitrack Motion Capturing system. After developing an offline-to-online Deep Reinforcement Learning algorithm using the previously collected data, the big moment arrived when the learned policy was deployed on the drone in the real world: “I was super amazed and happy when it worked in the very first experiment and started flying!”, Tiia said in her end-of-project presentation. Tiia will now continue her academic career by pursuing a Master of Science in Computational Science and Engineering at the TU Munich, Germany.
Combining Imitation Learning and Reinforcement Learning
Sergio Hernández is currently pursuing his Masters in Machine Learning, Data Science and Artificial Intelligence at Aalto University. For the summer of 2023 he joined the Intelligent Robotics lab as a summer intern working on new algorithms combining Imitation Learning and Deep Reinforcement Learning. This effort did not only require understanding and re-implementing complex state-of-the-art deep imitation learning techniques such as Co-Imitation, but also searching for relevant literature and deriving new mathematical frameworks from scratch. His algorithms were then tested in experiments run on our High-Perfomance Computing Infrastructure in the Lab and at Aalto University, such as the Triton Cluster.
“This project really opened my eyes to how different research work is from pure software engineering”, said Sergio, who worked previously as a software engineer at companies such as Rockstar Games and Cabify. Sergio will continue his work with the Intelligent Robotics Lab and use his results and algorithms developed during the summer for a future conference publication.