Biologically Inspired Artificial Intelligence for Adaptive and Efficient Robots
Contact person: Kai Olav Ellefsen
Keywords: Biologically Inspired Artificial Intelligence; Evolutionary Robotics-Continual Learning; Energy Efficient Machine Learning
Research group: Robotics and Intelligent Systems (ROBIN)
Department of Informatics, Department of Physics
While AI has made tremendous progress, animal bodies and brains are still far superior to AI algorithms and robots in many areas: They are more energy efficient, they need fewer examples to learn complex tasks, and they easily adapt to changing environments. Taking inspiration from biology, we aim to make robots and AI algorithms more robust, adaptive and energy efficient. Biological inspiration can take different forms: Evolutionary Robotics mimics natural selection to design efficient brains and bodies for robots. Biologically inspired learning can help make agents learn continuously in interaction with their environment. Combining inspiration from natural evolution and learning can allow us to optimize robot bodies and brains together, allowing robust and efficient control of robots that can adapt flexibly to new tasks and acquire new knowledge as they gain experience. Research proposals may span several methodological approaches and sources of inspiration from biology within the scope outlined here.
Methodological research topics:
- Reinforcement Learning
- Evolutionary Algorithms
- Meta-learning
Topics from natural sciences and technology:
- Neural learning mechanisms
- Energy efficient robot control and design
- Adaptivity in robotics Relevant partners:The Robotics and Intelligent Systems research group collaborates with several companies on applying robot technology.
External partners:
- Institute for Energy Technology (IFE)
- Kongsberg Maritime AS
- Statkraft AS
