Teacher

Francesco Visentin

Referent

Francesco Visentin

Credits

1.5

Language

English

Class attendance

Free Choice

Location

VERONA

Moodle Seminars 0

Learning objectives

This course aims to provide doctoral students with an interdisciplinary understanding of biologically inspired computation in the context of soft robotics. By the end of the course, students will be able to: understand the principles of morphological computation and embodied intelligence; describe the architecture and learning dynamics of Spiking Neural Networks (SNNs) and their relevance to neuromorphic engineering; explain the theory of Reservoir Computing and Echo State Networks, and their application to the control of compliant and continuum robots; understand the concept of Physical Reservoir Computing and how soft bodies can be exploited as computational substrates; critically analyze recent literature at the intersection of AI, neuroscience, and soft robotics; and identify open research challenges in the field.

Prerequisites and basic notions

Students are expected to have a basic background in robotics or mechanical/biomedical engineering, and familiarity with fundamental concepts of dynamical systems and control theory. Prior exposure to machine learning or neural networks is beneficial but not strictly required. A working knowledge of linear algebra and calculus is assumed.

Program

Topic 1 — From Nature to Engineering: Morphological Computation and Embodied Intelligence
Biological principles of movement and control. The concept of morphological computation. Embodied intelligence and brain-body co-design. Overview of soft robotics: materials, actuators, and fabrication.
Topic 2 — Soft Robotics: State of the Art
Soft and continuum robots: kinematics, mechanics, and modeling challenges. Distributed and embedded sensing. Overview of EIT-based sensing for soft structures. Current applications in medical and industrial robotics.
Topic 3 — Spiking Neural Networks (SNNs)
From rate coding to spike-based computation. Biological plausibility and energy efficiency. Neuron models: LIF, Izhikevich. Learning in SNNs: STDP and surrogate gradient methods. Neuromorphic hardware: Intel Loihi and BrainScaleS.
Topic 4 — Reservoir Computing: Theory and Intuition
Echo State Networks and Liquid State Machines. Stability conditions: echo state property and edge of chaos. Training only the readout layer. Comparison with recurrent neural networks.
Topic 5 — Physical Reservoir Computing
The soft body as a computational substrate. Key experiments and results (Nakajima et al.). Exploiting material dynamics for sensorimotor tasks. Design principles for physical reservoirs.
Topic 6 — Integrated Applications and Open Problems
Inverse kinematics of continuum robots via reservoir computing. Closed-loop control with SNN-based feedback. Embedded sensing and neuromorphic control. Open research challenges and future directions.

Didactic methods

The course is delivered through frontal lectures supported by slides and multimedia materials. Each lecture alternates between theoretical exposition and discussion of seminal and recent research papers. Students are encouraged to actively participate in critical analysis of the literature. Selected readings will be distributed in advance to foster informed discussion during class.

Learning assessment procedures

The exam consists of two parts. First, students are required to submit a short written report proposing a research line that integrates one or more topics covered in the course — such as SNNs, reservoir computing, or physical reservoir computing — within the context of their own doctoral research. The report should identify a specific open problem, motivate the proposed approach, and outline a preliminary methodology. Second, students will discuss their report in an individual oral session with the instructor, demonstrating their understanding of the course topics and their ability to critically connect them to their research area.

Assessment

Students will be assessed on the following criteria. For the written report: clarity and coherence of the proposed research line; demonstrated understanding of the course topics; originality and feasibility of the proposed approach; quality of scientific writing. For the oral discussion: ability to critically connect the course themes to their own doctoral research; depth of understanding of the underlying theoretical concepts; capacity to identify open problems and discuss potential solutions. Overall, particular value is placed on the student's ability to think across disciplinary boundaries and to envision concrete research applications of biologically inspired computational models in soft robotics.

Criteria for the composition of the final grade

The final grade is expressed as pass/fail. It is determined by a combined evaluation of the written report (60%) and the oral discussion (40%). A pass is awarded to students who demonstrate adequate understanding of the course topics, coherent integration of the covered themes within their research context, and satisfactory scientific communication skills in both written and oral form.

Scheduled Lessons

Sustainable Development Goals - SDGs

This initiative contributes to the achievement of the Sustainable Development Goals of the UN Agenda 2030. More information on sustainability