Trajectory Learning via Stable Dynamical Systems

In robotics Learning from Demonstration is the paradigm in which robots acquire new skills by learning to imitate an expert. Different families of techniques have been developed, each with its advantages and disadvantages (probabilistic or deterministic, interpretable or black-box, etc.).

One of these models is the Dynamic Movement Primitives framework. It consists of a perturbed spring-mass-damper system in which a forcing term is learned in order to model any desired behavior of the robot’s end-effector. The asymptotic stability of the dynamical system guarantees the convergence of the robot’s hand to any desired position, while the learned forcing function ensures that the executed movement will have a similar shape to the learned behavior.

The aim of the project is to implement the Dynamic Movement Primitives framework, in order to learn a three-dimensional trajectory and execute it while changing the final position. Background: Ordinary Differential Equations, Numerical Analysis, Python or Matlab

Instructor: Michele Ginesi (University of Verona)


  • Mariana Costa Villegas, Università di Padova
  • Venkata Sai Ganesh Ranganath Nalluri, Universität Koblenz
  • Alina Hermanchuk, Igor Sikorsky Kyiv Polytechnic Institute
  • Sebastiano Fregnan, University of Verona
  • Malina Elena Gradinaru, Università degli studi di Verona
  • Davide Coppola, University of Verona


Room T.02a, Veronetta, Istituto Ex-Orsoline, via Paradiso 6, 37129 Verona