Skip to main content

Neuromorphic Computing & Event Cameras

Duration: 55 min · Level: Advanced · Module: 9. Edge AI & On-Board Intelligence · Focus: neuromorphic, Loihi, event-cameras, low-power

Learning objectives

By the end of this lesson you will be able to explain and apply:

  • Intel Loihi 2 (2022)
  • Event cameras (DAVIS346, Prophesee EVK4)
  • SNN locomotion controller
  • Event camera + neuromorphic
  • Limitation

Why this matters​

Neuromorphic chips (Intel Loihi 2, IBM NorthPole) and event cameras offer a radically different compute paradigm: sparse, asynchronous, ultra-low-power processing inspired by biological neural circuits.

Overview​

Neuromorphic chips (Intel Loihi 2, IBM NorthPole) and event cameras offer a radically different compute paradigm: sparse, asynchronous, ultra-low-power processing inspired by biological neural circuits. For G1, they offer a path to always-on perception that consumes milliwatts rather than watts.

Key concepts​

Key idea

Intel Loihi 2 (2022): 1 million neurons, 120 million synapses per chip, 0.5W power; runs spiking neural networks (SNNs) 100× more energy-efficiently than equivalent GPU ops

  • Event cameras (DAVIS346, Prophesee EVK4): output asynchronous "events" (pixel, timestamp, polarity) when pixel brightness changes; μs latency, 120 dB dynamic range, no motion blur
  • SNN locomotion controller: Intel/ETH Zurich collaboration (2023) — running SNN locomotion policy on Loihi 2 for ANYmal; 75× lower energy than GPU equivalent; no meaningful performance loss
  • Event camera + neuromorphic: event stream → SNN on Loihi 2 → obstacle detection at 10,000 Hz update rate in <1mW; critical for detecting fast-moving hazards near humans
  • Limitation: programming SNNs requires specialized tools (NEST, PyNN, Intel's nxSDK); most researchers lack SNN expertise; steep learning curve
  • G1 application: Loihi 2 co-processor for always-on safety monitoring (detect human approach, collision prediction) while main Orin is in low-power sleep mode

Check your understanding​

Try to recall each answer before expanding it.

Q1. What do you know about Intel Loihi 2 (2022)?

1 million neurons, 120 million synapses per chip, 0.5W power; runs spiking neural networks (SNNs) 100× more energy-efficiently than equivalent GPU ops

Q2. What do you know about Event cameras (DAVIS346, Prophesee EVK4)?

output asynchronous "events" (pixel, timestamp, polarity) when pixel brightness changes; μs latency, 120 dB dynamic range, no motion blur

Q3. What do you know about SNN locomotion controller?

Intel/ETH Zurich collaboration (2023) — running SNN locomotion policy on Loihi 2 for ANYmal; 75× lower energy than GPU equivalent; no meaningful performance loss

Q4. What do you know about Event camera + neuromorphic?

event stream → SNN on Loihi 2 → obstacle detection at 10,000 Hz update rate in <1mW; critical for detecting fast-moving hazards near humans

Q5. What do you know about Limitation?

programming SNNs requires specialized tools (NEST, PyNN, Intel's nxSDK); most researchers lack SNN expertise; steep learning curve

References​

  • Intel Loihi 2: A New Generation of Neuromorphic Processor — Davies et al. (2022). IEEE Micro 2022
  • Neuromorphic Control of a Quadruped Robot — Müller-Cleve et al. (2023). Science Robotics 2023

← Previous: 9.2 Model Compression for Edge Deployment

Part of Module 9: Edge AI & On-Board Intelligence.