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

doi:10.3389/fnins.2020.00551

Cited by 22 publications

(17 citation statements)

References 68 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this "soft" neuromorphic approach, the front end clocked samples are converted to event-based representation by means of algorithms implemented in software [49][50][51] or embedded on Digital Signal Processors (DSPs) 52 or FPGAs 53,54 . The same approach is valuable also in other sensory modalities, such as proprioception 55,56 , to support the development of event-driven algorithms and validate their use in robotic applications. However, it is not optimal in terms of size, power, and latency.…”

Section: Box 1 | the Need For Adaptation In Roboticsmentioning

confidence: 99%

Embodied neuromorphic intelligence

2022

Self Cite

View full text Add to dashboard Cite

The design of robots that interact autonomously with the environment and exhibit complex behaviours is an open challenge that can benefit from understanding what makes living beings fit to act in the world. Neuromorphic engineering studies neural computational principles to develop technologies that can provide a computing substrate for building compact and low-power processing systems. We discuss why endowing robots with neuromorphic technologies – from perception to motor control – represents a promising approach for the creation of robots which can seamlessly integrate in society. We present initial attempts in this direction, highlight open challenges, and propose actions required to overcome current limitations.

show abstract

Section: Box 1 | the Need For Adaptation In Roboticsmentioning

confidence: 99%

Embodied neuromorphic intelligence

2022

Self Cite

View full text Add to dashboard Cite

show abstract

“…A stronger input results in a higher firing rate of the corresponding velocity neuron and a faster shift of the attractor-bump in the HD ring. As shown in [122], this velocity mapping can be precisely calibrated, achieving an error below 1 • . Drift in the pose representation is corrected when the robot recognizes a location that it has previously visited.…”

Section: Depending On the Sign And Magnitude Of The Error The Learned Map Is Updated And The Velocity Representation Is Calibratedmentioning

confidence: 98%

Advancing Neuromorphic Computing With Loihi: A Survey of Results and Outlook

et al. 2021

Self Cite

View full text Add to dashboard Cite

Deep artificial neural networks apply principles of the brain's information processing that led to breakthroughs in machine learning spanning many problem domains. Neuromorphic computing aims to take this a step further to chips more directly inspired by the form and function of biological neural circuits, so they can process new knowledge, adapt, behave, and learn in real time at low power levels. Despite several decades of research, until recently, very few published results have shown that today's neuromorphic chips can demonstrate quantitative computational value. This is now changing with the advent of Intel's Loihi, a neuromorphic research processor designed to support a broad range of spiking neural networks with sufficient scale, performance, and features to deliver competitive results compared to state-of-the-art contemporary computing architectures. This survey reviews results that are obtained to date with Loihi across the major algorithmic domains under study, including deep learning approaches and novel approaches that aim to more directly harness the key features of spike-based neuromorphic hardware. While conventional feedforward deep neural networks show modest if any benefit on Loihi, more brain-inspired networks using recurrence, precise spike-timing relationships, synaptic plasticity, stochasticity, and sparsity perform certain computation with orders of magnitude lower latency and energy compared to state-of-the-art conventional approaches. These compelling neuromorphic networks solve a diverse range of problems representative of brain-like computation, such as event-based

show abstract

“…Thus, in this work, the error signal integration was based on a path integration network developed for neuromorphic SLAM in [15], [28]. Here, two different levels of integration are introduced: (1) on the level of an individual spiking neuron and (2) on the level of a population of neurons.…”

Section: B the Spiking Neural Network (Snn) Pid 1) Snn Input Encodingmentioning

confidence: 99%

“…These properties make neuromorphic hardware an attractive platform for neurorobotics. Neuromorphic platforms were used in proof-of-concept experiments for, e.g., robot navigation, path planning, and SLAM [10]- [15]. More recently, several architectures for spike-based motor control were suggested, and some of them realized in neuromorphic hardware, showing promising results.…”

Section: Introductionmentioning

confidence: 99%

Event-based PID controller fully realized in neuromorphic hardware: a one DoF study

Stagsted¹,

Vitale

Renner

et al. 2020

2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Self Cite

View full text Add to dashboard Cite

Spiking Neuronal Networks (SNNs) realized inneuromorphic hardware lead to low-power and low-latencyneuronal computing architectures. Neuromorphic computingsystems are most efficient when all of perception, decision making , and motor control are seamlessly integrated into a singleneuronal architecture that can be realized on the neuromorphichardware. Many neuronal network architectures address theperception tasks, while work on neuronal motor controllersis scarce. Here, we present an improved implementation of aneuromorphic PID controller. The controller was realized onIntel's neuromorphic research chip Loihi and its performancetested on a drone, constrained to rotate on a single axis. TheSNN controller is built using neuronal populations, in whicha single spike carries information about sensed and controlsignals. Neuronal arrays perform computation on such sparserepresentations to calculate the proportional, derivative, andintegral terms. The SNN PID controller is compared to a PIDcontroller, implemented in software, and achieves a comparableperformance, paving the way to a fully neuromorphic systemin which perception, planning, and control are realized in anon-chip SNN.

show abstract

An On-chip Spiking Neural Network for Estimation of the Head Pose of the iCub Robot

Cited by 22 publications

References 68 publications

Embodied neuromorphic intelligence

Embodied neuromorphic intelligence

Advancing Neuromorphic Computing With Loihi: A Survey of Results and Outlook

Event-based PID controller fully realized in neuromorphic hardware: a one DoF study

Contact Info

Product

Resources

About