Project

Back to overview

Embodied Long-Term Memory for Cognitive Neuromorphic Agents

Applicant Sandamirskaya Yulia
Number 168183
Funding scheme Ambizione
Research institution Institut für Neuroinformatik Universität Zürich Irchel und ETH Zürich
Institution of higher education University of Zurich - ZH
Main discipline Information Technology
Start/End 01.06.2017 - 30.09.2020
Approved amount 577'041.00
Show all

All Disciplines (2)

Discipline
Information Technology
Other disciplines of Engineering Sciences

Keywords (5)

Cognitive systems; Neurorobotics; Embodied Cognition; Neuromorphic Engineering; Neural dynamics

Lay Summary (German)

Lead
Das Langzeitgedächtnis bildet eine Grundlage für das intelligente Verhalten und ist ein wichtiger Ausgang von Lernprozessen. Die elementaren neuronalen Lernmechanismen sind gut erforscht und können mathematisch erfasst und auf Computern simuliert werden. Es ist, im Gegensatz, wenig verstanden, wie das Langzeitgedächtnis durch Lernen entsteht, erhalten bleibt, und immer mit neuem Wissen und Können erweitert wird. Das ELMA Projekt wird die Formation des Langzeitgedächtnises durch das Bauen von künstlichen neuromorphen Agenten untersuchen, welche durch die Interaktion mit der Umwelt neue Repräsentationen im neuronalen Langzeitgedächtnis aufbauen.
Lay summary

Inhalt und Ziel des Forschungsprojekts

Das Ziel des Projekts ist es, das Langzeitgedachtnis fur Objekte, Szenen, Sequenzen, und Aktionen zu untersuchen. Die neuronalen Modelle von Lernprozessen werden in den an der UZH entwickelten neuromorphen Chips implementiert, welche realistische neuronale Dynamiken in Echtzeit direkt in elektronischen Schaltkreisen nachbilden. Um die Formung des Langzeitgedachtnises durch Interaktion mit der Umwelt zu ermoglichen, werden die neuronalen Modelle an die Sensoren und Motoren von robotischen Agenten gekoppelt. In so erbauten “embodied” Agenten werden die Prozesse der Gedachtnisaufbau in geschlossenem Verhaltenskreis untersucht. Die neu entwickelten Modelle werden mit experimentellen Erkenntnissen verglichen. 

Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts 

Das Verstandnis von Langzeitgedachtnis fuhrt nicht nur zum besseren Verstandnis von vielen kognitiven Prozessen, z.B. Lernen, Entscheidungsfindung, oder Handlungsplanung, sondern auch zu besseren kunstlich intelligenten Systemen, welche in echter Umwelt agieren und lernen konnen. Solche Systeme werden durch Einsatz von neuromorphen Technologien weniger Energie verbrauchen als herkommliche kunstliche neuronale Netze, und auch schnell von eigener Erfahrung lernen konnen. Sie konnten in, z.B., prothetischen Geraten, oder auch in Sicherheitssystemen oder Assistenzrobotern eingesetzt werden. 

Keywords

Neuromorphes Engineering; neuronale Dynamiken; Langzeitgedachtnis; autonomes Lernen; dynamische neuronal Felder; kognitive Robotik; Neurorobotik. 

Direct link to Lay Summary Last update: 04.05.2017

Responsible applicant and co-applicants

Employees

Publications

Publication
Visual Pattern Recognition with on On-Chip Learning: Towards a Fully Neuromorphic Approach
Baumgartner Sandro, Renner Alpha, Kreiser Raphaela, Liang Dongchen, Indiveri Giacomo, Sandamirskaya Yulia (2020), Visual Pattern Recognition with on On-Chip Learning: Towards a Fully Neuromorphic Approach, in 2020 IEEE International Symposium on Circuits and Systems (ISCAS), SevillaIEEE, Virtual.
A Digital Multiplier-less Neuromorphic Model for Learning a Context-Dependent Task
Asgari Hajar, Maybodi Babak Mazloom-Nezhad, Kreiser Raphaela, Sandamirskaya Yulia (2020), A Digital Multiplier-less Neuromorphic Model for Learning a Context-Dependent Task, in 2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), Genova, ItalyIEEE, Genova, Italy.
Towards neuromorphic control: A spiking neural network based PID controller for UAV
Stagsted Rasmus, Vitale Antonio, Binz Jonas, Renner Alpha, Bonde Larsen Leon, Sandamirskaya Yulia (2020), Towards neuromorphic control: A spiking neural network based PID controller for UAV, in Robotics: Science and Systems 2020, Robotics: Science and Systems Foundation, Virtual.
An On-chip Spiking Neural Network for Estimation of the Head Pose of the iCub Robot
Kreiser Raphaela, Renner Alpha, Leite Vanessa R. C., Serhan Baris, Bartolozzi Chiara, Glover Arren, Sandamirskaya Yulia (2020), An On-chip Spiking Neural Network for Estimation of the Head Pose of the iCub Robot, in Frontiers in Neuroscience, 14, 551.
Error estimation and correction in a spiking neural network for map formation in neuromorphic hardware
Kreiser Raphaela, Waibel Gabriel, Armengol Nuria, Renner Alpha, Sandamirskaya Yulia (2020), Error estimation and correction in a spiking neural network for map formation in neuromorphic hardware, in 2020 IEEE International Conference on Robotics and Automation (ICRA), Paris, FranceIEEE, Paris, France.
Neural State Machines for Robust Learning and Control of Neuromorphic Agents
Liang Dongchen, Kreiser Raphaela, Nielsen Carsten, Qiao Ning, Sandamirskaya Yulia, Indiveri Giacomo (2019), Neural State Machines for Robust Learning and Control of Neuromorphic Agents, in IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 9(4), 679-689.
Autonomous Sequence Generation for a Neural Dynamic Robot: Scene Perception, Serial Order, and Object-Oriented Movement
Tekülve Jan, Fois Adrien, Sandamirskaya Yulia, Schöner Gregor (2019), Autonomous Sequence Generation for a Neural Dynamic Robot: Scene Perception, Serial Order, and Object-Oriented Movement, in Frontiers in Neurorobotics, 13, 95.
Robust Learning and Recognition of Visual Patterns in Neuromorphic Electronic Agents
Liang Dongchen, Kreiser Raphaela, Nielsen Carsten, Qiao Ning, Sandamirskaya Yulia, Indiveri Giacomo (2019), Robust Learning and Recognition of Visual Patterns in Neuromorphic Electronic Agents, in 2019 IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), Hsinchu, TaiwanIEEE, Hsinchu, Taiwan.
Event-Based Attention and Tracking on Neuromorphic Hardware
RennerAlpha, EvanusaMatthew, SandamirskayaYulia (2019), Event-Based Attention and Tracking on Neuromorphic Hardware, in The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Long Beach, CAIEEE Xplore, online.
Robust Learning and Recognition of Visual Patterns in Neuromorphic Electronic Agents
LiangDongcheng, KreiserRaphaela, SandamirskayaYulia, IndiveriGiacomo (2019), Robust Learning and Recognition of Visual Patterns in Neuromorphic Electronic Agents, in 1st IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), Hsinchu, TaiwanIEEE, online.
Learning Temporal Intervals in Neural Dynamics
Duran Boris, Sandamirskaya Yulia (2018), Learning Temporal Intervals in Neural Dynamics, in IEEE Transactions on Cognitive and Developmental Systems, 10(2), 359-372.
A Neuromorphic Approach to Path Integration: A Head-Direction Spiking Neural Network with Vision-driven Reset
Kreiser Raphaela, Cartiglia Matteo, Martel Julien N.P., Conradt Jorg, Sandamirskaya Yulia (2018), A Neuromorphic Approach to Path Integration: A Head-Direction Spiking Neural Network with Vision-driven Reset, in 2018 IEEE International Symposium on Circuits and Systems (ISCAS), Florence, ItalyIEEE, IEEE Xplore.
An Active Approach to Solving the Stereo Matching Problem using Event-Based Sensors
Martel Julien N. P., Muller Jonathan, Conradt Jorg, Sandamirskaya Yulia (2018), An Active Approach to Solving the Stereo Matching Problem using Event-Based Sensors, in 2018 IEEE International Symposium on Circuits and Systems (ISCAS), Florence, ItalyIEEE, IEEE Xplore.
Real-Time Depth From Focus on a Programmable Focal Plane Processor
Martel Julien N. P., Muller Lorenz K., Carey Stephen J., Muller Jonathan, Sandamirskaya Yulia, Dudek Piotr (2018), Real-Time Depth From Focus on a Programmable Focal Plane Processor, in IEEE Transactions on Circuits and Systems I: Regular Papers, 65(3), 925-934.
Learning and adaptation: neural and behavioural mechanisms behind behaviour change
Lowe Robert, Sandamirskaya Yulia (2018), Learning and adaptation: neural and behavioural mechanisms behind behaviour change, in Connection Science, 30(1), 1-4.
Dynamic Neural Fields with Intrinsic Plasticity
Strub Claudius, Schöner Gregor, Wörgötter Florentin, Sandamirskaya Yulia (2017), Dynamic Neural Fields with Intrinsic Plasticity, in Frontiers in Computational Neuroscience, 11, 74.
Adaptive motor control and learning in a spiking neural network realised on a mixed-signal neuromorphic processor
GlatzSebastian, MartelJulien N. P., KreiserRaphaela, QiaoNing, SandamirskayaYulia, Adaptive motor control and learning in a spiking neural network realised on a mixed-signal neuromorphic processor, in IEEE International Conference on Robotcs and Automation (ICRA), IEEE Xplore, Online.
The importance of space and time for signal processing in neuromorphic agents
IndiveriGiacomo, SandamirskayaYulia, The importance of space and time for signal processing in neuromorphic agents, in IEEE Signal Processing Magazine.

Collaboration

Group / person Country
Types of collaboration
Florentin Wörgötter, Computational Neurosciences, University of Göttingen Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Fatih Yanik, Neurotechnology Group, INI at UZH/ETH Zurich Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Abhishek Banerjee, Laboratory of Neural Circuit Dynamics, HIFO, UZH Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Jörg Conradt, Neuroscientific System Theory, TU Munich, Germany Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
Piotr Dudek, Circuits and Systems group, University of Manchester Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Robust AI for Neurorobotics Talk given at a conference Error-driven learning for self-calibration in a neuromorphic path integration system 26.08.2019 Edinburgh, Great Britain and Northern Ireland Sandamirskaya Yulia;
IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019 Talk given at a conference Neuromorphic Computing: towards event-based (cognitive) sensing and control 17.06.2019 Long Beach, CA, United States of America Sandamirskaya Yulia;
IEEE Symposium for Circuits and Systems, ISCAS Talk given at a conference (1) Live Demonstration: An Active System for Depth Reconstruction using Event-Based Sensors; Talk: A Neuromorphic Approach to Path Integration: a Head Direction Spiking Neural Network with Vision-Driven Reset 27.05.2019 Florence, Italy Sandamirskaya Yulia;
IEEE International Conference on Robotics and Automation (ICRA) 2019 Poster Adaptive motor control and learning in a spiking neural network realised on a mixed-signal neuromorphic processor 20.05.2019 Montreal, Canada Sandamirskaya Yulia;
Neural Control of Movement (NCM), Satellite workshop “Predictive coding and active inference to know and explore the world” Talk given at a conference Neuronal attractor dynamics: coupling perception, movement generation, and learning 23.04.2019 Toyama, Japan Sandamirskaya Yulia;
Dagstuhl Seminar 19152 “Emerging Hardware Techniques and EDA Methodologies for Neuromorphic Computing” Talk given at a conference Neuromorphic Architectures for Robots 07.04.2019 Schloss Dagstuhl, Germany Sandamirskaya Yulia;
Neuro-Inspired Computational Elements (NICE) Workshop Talk given at a conference Attractor Dynamics and Embodiment of Neural Computing 26.03.2019 Albany, NY, United States of America Sandamirskaya Yulia;
Redwood Seminar; Berkeley Neuroscience Individual talk Dynamic Neural Fields: The Embodiment of Neural Computation 05.02.2019 Berkeley, CA, United States of America Sandamirskaya Yulia;
Cognitive Computing: Merging Concepts with Hardware Talk given at a conference Embodied Neuromorphic Cognition and Learning 18.12.2018 Hannover, Germany Sandamirskaya Yulia;
Symposium “Controlling Behavior in Animals and Robots" Talk given at a conference Neuronal Dynamics of Action Representation, Execution, and Learning 03.12.2018 EPFL, Lausanne, Switzerland Sandamirskaya Yulia;
IROS 2018: IEEE/RSJ International Conference on Intelligent Robots and Systems Talk given at a conference Pose Estimation and Map Formation with Spiking Neural Networks: towards Neuromorphic SLAM 01.10.2018 Madrid, Spain Sandamirskaya Yulia;
Bernstein Conference 2018; Workshop: Learning and recalling sequences of actions at the neuronal level and beyond Talk given at a conference Neural dynamics of autonomous sequence learning 25.09.2018 Berlin, Germany Renner Alpha; Sandamirskaya Yulia;
ZNZ Symposium 2018 Poster Sensory processing and Motor Control for Neuromorphic Robotics 13.09.2018 Zurich, Switzerland Sandamirskaya Yulia; Renner Alpha;
Telluride 2018 Neuromorphic Cognition Engineering Workshop Talk given at a conference Tutorial: Neural Fields and Neuromorphic Hardware 01.07.2018 Telluride, CO, United States of America Sandamirskaya Yulia;
Dynamic Field Theory: Expanding the Field Talk given at a conference Neuronal basis of cognition and autonomy of acting, learning, and development 03.06.2018 Norwich, Great Britain and Northern Ireland Sandamirskaya Yulia;
Morphological and Embodied Computing. Symposium on Theory and Applications. Talk given at a conference Neuromorphic Cognitive Agents 07.05.2018 Gothenburg, Sweden Sandamirskaya Yulia;
Capo Caccia Cognitive Neuromorphic Engineering Workshop Talk given at a conference "Neuromorphic Robots" Workshop and Lectures 23.04.2018 Alghero, Hotel Dei Pini, Italy Renner Alpha; Sandamirskaya Yulia;
EITN "From Neuroscience to Machine Learning" Talk given at a conference Beyond predictive coding: deciding what and when to (un)learn in a closed sensorimotor loop 12.03.2018 Paris, France Sandamirskaya Yulia;
EITN Workshop "Cortical codes" Talk given at a conference Neural Dynamics of Behavior Generation and Autonomous Learning 06.02.2018 Paris, France Sandamirskaya Yulia;
The Science of Cognition: Brain, Body, Language, and Technologies, AISC Talk given at a conference Neural Dynamics of Cognition 14.12.2017 Bologna, Italy Sandamirskaya Yulia;
IEEE/RSJ International Conference on Intelligent Robots and Systems Talk given at a conference Architecture for Action Parsing using SCAMP 23.09.2017 Vancouver, United States of America Sandamirskaya Yulia;
ZNZ Symposium 2017 Poster The sense of space: spatial representations in biological and artificial spiking neural networks 14.09.2017 Zurich, Switzerland Sandamirskaya Yulia;
ZNZ Symposium 2017 Talk given at a conference Exploring neuronal basis of cognition using neuromorphic devices in a closed sensory-motor loop 14.09.2017 Zurich, Switzerland Sandamirskaya Yulia;
Robotics: Science and Systems Talk given at a conference A neuromorphic controller for a robotic vehicle equipped with a dynamic vision sensor 12.07.2017 Boston, United States of America Sandamirskaya Yulia;


Self-organised

Title Date Place
ICRA 2020 Workshop "Unconventional Sensing in Robotics" 30.05.2020 Paris (virtual), France
First NEUROTECH Forum 04.11.2019 Leuven, Netherlands

Knowledge transfer events

Active participation

Title Type of contribution Date Place Persons involved
Kader Referat Kantonpolizei Zürich Talk 09.07.2019 Zurich, Switzerland Sandamirskaya Yulia;
Skriptorium Gespräch Talk 07.06.2019 St. Gallen, Stiftungsbibliothek, Switzerland Sandamirskaya Yulia;
Visit Intel Labs, Portland; Wiring-up the Neuromorphic Brain: Neuro-Dynamic Architectures for Embodied Cognition Talk 06.02.2019 Portland, Oregon, United States of America Sandamirskaya Yulia;
CSEM Seminar Talk 22.06.2017 Zurich, Switzerland Sandamirskaya Yulia;


Communication with the public

Communication Title Media Place Year
Media relations: print media, online media Talk im Turm "Was uns die Zukunft bringt" News UZH German-speaking Switzerland 2019
Media relations: radio, television "Roboter mischen Davos auf" N-TV International 2018
Media relations: print media, online media "Sie entwickelt unsere Zukunft" Tagesanzeige German-speaking Switzerland 2018
Talks/events/exhibitions 100 Ways of Thinking (Contribution: Visual Illusion Installation; Video Interview) German-speaking Switzerland 2018
Media relations: radio, television Aeschbacher SRF 1 German-speaking Switzerland 2018
Talks/events/exhibitions Davos World Economic Forum, ETH@Davos: "Rethinking Intelligence" International 2018
Media relations: print media, online media Interdisziplinarität fördert Intelligenz Schaffhauser Nachrichten German-speaking Switzerland 2018
Talks/events/exhibitions National Future Day; INI Tour, Demos, Intro German-speaking Switzerland 2018
Talks/events/exhibitions Neuronale Netze: Vom Menschlichen Hirn zu Künstlicher Intelligenz und Autonomen Robotern German-speaking Switzerland 2018
Media relations: print media, online media Sie entwickelt unsere Zukunft Tagesanzeige German-speaking Switzerland 2018
Media relations: print media, online media Wir entscheiden, was Roboter können Schaffhauser Nachrichten German-speaking Switzerland 2018
Talks/events/exhibitions Neuromorphic Robots German-speaking Switzerland 2017
Talks/events/exhibitions Scientifica 2017 ("Neuromorphe Computer: vom Menschlichen Hirn zu kognitiven Robotern) German-speaking Switzerland 2017

Awards

Title Year
Best Demo Prize (2. Prize) IEEE Symposium on Circuits and Systems 2018
Best Demo Award (2. Prize) IEEE Symposium on Circuits and Systems 2017

Abstract

The goal of this project is to enhance a new type of intelligent sensorimotor systems that are currently being developed by the applicant -- the neuromorphic embodied cognitive systems -- with long-term memory. This goal has a theoretical and a technological components. In the theory part, the mechanisms of formation of long-term memories in neuronal systems will be realised in computational architectures that are embodied, i.e. receive inputs from sensors and can generate actions in an environment. Such long-term memories form the basis of cognition and intelligent behavior by storing representations of objects, actions, episodes, and temporal sequences. The technological goal of the project is to create cognitive neuromorphic robotic agents that can act and perceive in the real world and build representations in an interactive process with the environment. The project aims to enable interaction of the theory- and technology-driven research. By creating a technology for realisation of neuronal cognitive architectures in robotic agents, the project will enable testing models and generating new hypothesis for neuroscience experiments in a closed-loop behavioral setting. On the other hand, the theoretical insights gained when developing computational neuronal models capable to generate long-term memory for objects, actions, and spatio-temporal relations between them can drive development of new algorithms for solving tasks that involve sensorimotor interactions and require adaptivity and learning in new environments. To achieve this goal, I will use the computational framework of attractor dynamics. This neural-dynamic framework has been used in the past both to account for activity of neuronal populations in form of Dynamic Neural Fields (DNF) models and to explain the sensorimotor and developmental basis of cognitive behavior in the embodied cognition paradigm. In this project, the DNF framework will be extended with architectures for forming and storing long-term memories, developed in collaboration with experimental neuroscientists, who probe formation of episodic and sequential memories in rodents and formation of representations of songs in birds. The modelling work will be based on a body of work on embodied cognition, computational neuroscience, artificial intelligence, and cognitive robotics. The obtained computational models will be realised in spiking neuromorphic VLSI technology, which is low-power and efficient for implementing neuronal architectures. These neuromorphic cognitive hardware will be interfaced to sensors and motors of robotic vehicles, which will generate action sequences and form long-term memories of the observed environmental contingencies. Adaptivity, learning, and long-term memory formation of the cognitive robotic agents will be probed in benchmark scenarios. This project will play a key role in moving neuromorphic devices towards application in real-world scenarios: servailliance, robotics, and smart environments and will have an impact in the research fields of machine learning, cognitive robotics, and computational neuroscience.
-