Join us Tuesday, October 18, 2022 @ 9:00-10:30am PDT / 18:00-19:30 CET for another great INRC Forum, this time featuring an update on robotics at Intel and fortiss!

Agenda

• @Yulia Sandamirskaya (Deactivated) will share an update on the latest robotics R&D in Intel Neuromorphic Computing Lab and outlook to new capabilities in Lava.

• Research Talk by INRC member @Evan Eames and @amaya of fortiss.

• @Mathis Richter hosts Lava Open-Door / Community Meeting. Ask questions and get feedback from Lava developers.

Open

 

Spiking Reinforcement Learning for Force Feedback Based Robotic Object Insertion on Loihi

Abstract: Robotic object insertion is a classic task in robotics and has been approached with a variety of different learning algorithms and machine learning frameworks. Here we present the first successful implementation of the peg-in-hole task on neuromorphic architecture without vision. The task was first trained within the Neurorobotics Platform using the Spiking Reinforcement Learning technique proposed in Tang 2020. The trained policy was then moved onto neuromorphic hardware, specifically the Intel Loihi Research Chip. Finally, the network was connected to a real KUKA robotic arm with a force-torque sensor implemented on the end effector, allowing the neuromorphic hardware to control the movement of the arm, and guide it through the insertion, in real time. Domain randomization and system identification were used for closing the Sim2Real gap.

Open

Bio: Evan did his studies in astrophysics, specifically early universe cosmology. After his PhD he switched into ML, working for three years as a Data Science Consultant specializing in industrial ML solutions. On feeling that he’d arrived at the “ceiling” of how far one can go with ML within industry Evan switched back into research, this time taking up a postdoctoral position in Neuromorphic Computing within fortiss – the Bavarian State Research Institute. Thus far his work has focused mainly on creating real Neurorobotic use cases, and also incorporating elements of event-based vision. He also enjoys collecting random facts for pub trivia nights, and writing non-fiction in cafés.

Open

Camilo started his career studying mechatronics with a focus on robotics and automation. After finishing his engineering studies and gathering some experience in industry, he decided to follow his passion by doing an MSc. in Robotics, Cognition, and Intelligence. During his master’s degree he became involved in bio-inspired technologies and brain-inspired AI, and accordingly, he wrote his master’s thesis alongside the Neuromorphic Computing competence field within fortiss – the Bavarian state research institute for software-intensive systems and AI. Now he is a full-time researcher at fortiss, and his fields of research include event-based vision, neuromorphic computing, reinforcement learning, and robotics.

How to join:

INRC Members can find the meeting link on INRC Fall Forum Schedule. If you are interested in becoming a member, join the INRC.