From biology to robotics and back by Manos Angelidis

124

What makes our brains so efficient in computing and adapting to their environment? Why are animals so efficient at performing simple tasks that robots still find extremely difficult to perform? How can we learn from the interaction between neuroscience and robotics, shed light to complex biological phenomena and at the same time design novel robots?

Biologically inspired robotics has been a long standing research field that draws inspiration from biomechanics, neuroscience and robotics, and aims in both explaining the mechanisms with which organisms interact with their environment, learn and adapt, and at the same time create better algorithms and robots that can perform at a near to human cognition level.

Deep Learning has been very successful in providing near -or even above- human level intelligence performance in certain problems, but has no biological plausibility and often needs an enormous amount of data and processing power to train properly.

We will discuss about a category of methods and algorithms that can pose as an alternative and complement to DL, especially for applications which are energy and computationally intense. We will explore the methods behind biologically inspired robotics, and see how the insights that we can gain from robotics can provide useful feedback about our hypotheses of human cognition.

Slides

You can download the slides from SlideShare:

About Manos Angelidis

Manos Angelidis holds a Masters degree in Mechanical Engineering, and a Masters degree in Biomedical Engineering from the National Technical University of Athens. He spent 3 years as a Biomedical Engineering researcher in Athens on the topic of computational biomechanics and fluid dynamics. For the last 3 years he has been working at fortiss GmbH, a research institute for software and AI, also doing PhD research at the department of Informatics of the Technical University of Munich on the topic of AI and Robotics. He has been a software architect, lead senior software engineer and scrum Master in the distributed team developing the Neurorobitcs Platform within the Human Brain Project. He has expertise in simulated physics and Neuromorphic Computing, and specializes in C++, Javascript, Python, and Agile Methodologies.