← Back ICRA 2026

Optical Flow Estimation Using Speck Neuromorphic Hardware

Manupriya Singh, Dequan Ou, Jesse J. Hagenaars, Guido C. H. E. de Croon

PDF

AI summary

Key figure (auto-extracted from paper)

A hybrid spiking-artificial neural network on the compact SynSense Speck chip enables real-time dense optical flow estimation and stable closed-loop drone flight with twice the speed of traditional ANN baselines.

Neuromorphic computing Optical flow SynSense Speck Event-based vision Autonomous drones Hybrid SNN-ANN

Problem

Resource-constrained micro aerial vehicles require low-latency, energy-efficient perception for real-time autonomy, but conventional dense optical flow algorithms are too computationally heavy and fully asynchronous neuromorphic pipelines remain unproven on compact chips like Speck.

Approach

The authors convert a lightweight convolutional encoder into a spiking neural network for the Speck chip, while offloading recurrent memory and decoding layers to a companion computer, creating a hybrid pipeline that estimates dense flow from asynchronous event-camera data.

Key results

First closed-loop drone control using dense optical flow on the Speck chip
Real-time dense flow inference preserving ANN accuracy
~2× faster inference latency than ANN-only baseline at identical power
Successful stable indoor hovering and forward flight driven by flow cues

Why it matters

Proves that compact neuromorphic hardware can deliver ultra-efficient, low-latency perception for practical autonomous flight in power- and compute-constrained environments.

Abstract

Neuromorphic hardware and spiking neural net- works (SNNs) offer a bio-inspired path to low-latency, energy- efficient computation by emulating the brain’s asynchronous, spike-based processing. This is particularly attractive for resource-constrained robots that are tightly limited in size, weight, and power. We propose a neuromorphic approach to real-time optical flow estimation tailored to the SynSense Speck system-on-chip, which integrates a Dynamic Vision Sensor (DVS) with a neuromorphic processor. Our inference architecture combines spiking and artificial neural layers in a hybrid SNN–ANN framework, enabling the use of Speck to perform regression for closed-loop drone control, an application not previously demonstrated on this chip. Despite its compact form factor, the system produces dense flow in real time and achieves stable indoor hover and forward flight using flow- based control. The hybrid pipeline runs ∼2× faster than an ANN-only baseline at identical power, highlighting the promise of neuromorphic sensing and processing for ultra-efficient autonomous flight in real-world scenarios. Code and data are available at: https://mavlab.tudelft.nl/speck-optical-flow

Index terms

Sensor-based Control Computer Vision for Automation Aerial Systems: Perception and Autonomy