Netherlands, Nijmegen

Biomimetic Robot Navigation

when 8 July 2019 - 12 July 2019
language English
duration 1 week
credits 2 ECTS
fee EUR 780

Would you like your mobile robot to navigate autonomously just like animals find their way in an unknown environment? Then come and join us in this Summer School to build your own robocar and teach it how to navigate without a map.

The next new car you will own is likely to drive itself, at least under certain conditions like on a motorway. If you wait for 3-8 years, your new car might not even have a driving wheel, as Level 5 automation will enable robocars to navigate any road under any condition without human supervision. However, even then, the robocar will fail to match the navigational prowess of a rodent.

Rodent navigation is driven by novelty seeking, as robocars avoid collisions. While rodents navigate any territory, independent from whether they know the environment, robocars need a map to identify a path to travel. Thus if a robocar does not know where it is and where it needs to go, it cannot navigate its environment.

One process that unifies rodent and robocar (or any robotic) navigation is the necessity to use sensory inputs to generate motor commands, enabling navigation; this sensorimotor computation is at the core of decision-making in biological and artificial brains. What makes animal sensorimotor decision-making different to that of robots’ is the ability of the animal brains to infer the state of the world despite that sensory information might be incomplete, unreliable or otherwise noisy, to enable decision-making. Ever expanding variety, increased spatial resolution, improved sensitivity of sensors and the intelligent city/mesh-sensor infrastructure available for robotic sensing will overcome biological systems’ limitations. However, the requirement to create contextual information from high-dimensional input places a high computational cost for robotic decision-making. The brain solves this problem via sensory fusion and inference to create abstract representations of its environment to enable motor planning and navigation.

In this Radboud Summer School, we will take advantage of what Systems Neuroscience has unraveled about how animals navigate their environments and deploy them in robotic cars to perform autonomous navigation without a map.

The Summer School consists of lectures and hands-on laboratory practicals. Morning sessions are devoted to classroom lectures that will provide you with in-depth information about biomimetic animal navigation and control algorithms. In the afternoon you will have hands-on practicals during which you will build your own robotic car and deploy navigation algorithms to get ready for a robotic navigation competition on Friday afternoon.

Course leader

Tansu Celikel
Proffesor and Chair of Neurophysiology & Department Head
Department of Neurophysiology
Donders Institute

Ronny Hartanto
Professor
Faculty of Technology and Bionics
Hochschule Rhein-Waal

Target group

Advanced bachelor, master, PhD, post-doc and professional. The course is designed for Systems Neuroscientists, Roboticists, Computer Scientists and anyone with an advanced STEM education who is interested in robotic navigation.

Course aim

After this course you are able to:
1.Describe how brain performs sensorimotor computations,
2.Build your own robotic car and program it,
3.Reproduce the brain’s sensorimotor computations in a robotic agent,
4.Generalize the robotic control algorithms across contexts.

Fee info

EUR 780: Normal fee

Scholarships

€ 702 early bird discount – deadline 1 March 2019 (10%)
€ 663 partner + RU discount (15%)
€ 585 early bird + partner + RU discount (25%)