Event cameras are finally getting the FPGA treatment they deserved a decade ago
Two papers this week show visual odometry and event-based sensing moving toward hardware that actually makes sense for robots.
画像クレジット: Lottie animation by Centre Robotics (LottieFiles Free, used with credit). · source
Less than 2 kilobytes of storage. That's what a new FPGA-based event camera processor needs to estimate motion velocity, according to researchers who just posted their work on arXiv. When I was at Kuka, we were throwing megabytes at simpler problems and still missing deadlines.
Look, here's the thing: event cameras have been the perpetual "next big thing" in robotics vision for about fifteen years now. I remember seeing early prototypes at Automatica back in 2012, maybe 2013, and thinking we'd all be using them within five years. We weren't. The sensors worked fine. The processing didn't.
The hardware problem nobody wanted to solve
Event cameras output asynchronous streams of brightness changes rather than full frames. Great for latency, terrible for conventional processors that expect neat rectangular images at fixed intervals. Most research groups just threw GPUs at the problem and called it a day. Fine for a lab demo, useless for a drone or a mobile manipulator where every watt counts.
This new EventShiftFlow approach does something I genuinely didn't expect to see: it ditches floating-point math entirely. No DSP blocks. No dividers. Just shift registers, counters, and comparators. The kind of logic you could have implemented on hardware from 2008 if anyone had bothered to think it through.
The tradeoff is you get sparse, quantized velocity estimates rather than dense optical flow. The researchers are upfront about this (refreshingly so). They're not claiming to solve every vision problem. They're targeting reactive obstacle avoidance on platforms where size, weight, and power actually matter. That's a real engineering constraint, not an academic abstraction.
関連記事
More in Industrial
The acquisition signals Autodesk's push beyond CAD software into the messy reality of keeping physical assets running, though whether this creates genuine synergies or just a larger software bundle remains to be seen.
Aisha Patel · 6 hours ago · 8 min
More than you'd think, actually. Musk's IPO filing has some interesting implications for industrial automation.
Robert "Bob" Macintosh · 8 hours ago · 3 min
The global rush toward generative AI is pulling venture dollars away from emerging markets, and African robotics companies are feeling the pinch.
Aisha Patel · 14 hours ago · 6 min
Two days of demos, talks, and networking won't answer the hard questions about where this industry is actually headed.


