Mohammad Ali Zamani

Language-modulated safer actions

Principle Supervisor:
Prof. Dr. Stefan Wermter
Universität Hamburg

Collaboration partners:

  • University of Plymouth
  • Fraunhofer-Institut für Produktionstechnik und Automatisierung IPA

Competence Area: Situation


Objectives

To develop a neuro-inspired model (deep learning in convolution networks) of spoken language understanding for modulating actions. Humans often use short warnings to communicate a threatening situation to another human. If robots are to interact with humans in safe open spaces they will have to be able to listen to those warnings that humans naturally generate and adapt their behaviour accordingly.


Expected Results

The result will be an auditory-guided robot which can protect its body or change its current behaviour based on auditory warnings. We will study how the auditory input (e.g. sound, word) can trigger actions from simple actions toward action sequences and will develop an architecture for a robot that learns safe movements guided by auditory cues based on previously learned experiences.