The failure is not always visible
Jamming announces itself. Spoofing does not. The receiver reports a confident position that is wrong, and everything downstream believes it.
EvioNav
Event vision navigation for harsh environments.
Position for drones and autonomous vehicles when satellites cannot be trusted. EvioNav's event-based camera keeps the fixes coming in darkness, glare, and motion blur.
- 01 GNSS is denied
Jamming or spoofing takes the satellite fix away.
- 02 The vision camera fades
Darkness, glare and motion blur leave it nothing to match.
- 03 The event camera takes over
Position fixes keep flowing to the vehicle control system.
The problem
GNSS fails. The vehicle still needs to know where it is.
Jamming and spoofing are no longer exotic. The signal can drop out, or it can quietly lie. Either way the vehicle has to know where it is and where it is going, well enough to find a known site, in darkness, backlight, cloud, motion and vibration.
The IMU only buys seconds
Inertial dead reckoning bridges short gaps. Over minutes the error compounds with nothing left to correct it, and the estimate stops being worth trusting.
The camera goes blind first
The fallback is to look at the ground. Night, low sun and hard motion are exactly when a vision camera stops returning an image worth matching.
Turning back has a price
Aborting is the honest answer today. It is also the mission not flown, the site not reached, the cargo still on the ground.
What EvioNav is
Position fixes that survive when the regular vision camera goes blind.
Vision navigation without GNSS already works: a vision camera matches the ground, an IMU bridges the gaps. It works, up to a limit. EvioNav does not replace any of that.
It is the event-camera layer of that stack. Event pixels fire on change, not on exposure, so fixes keep arriving in the darkness, glare and motion that leave a vision camera with nothing to match.
A layer, not a replacement
It sits alongside the vision camera and the IMU. While they hold, nothing changes. When they stop, the fixes keep coming.
Availability, not extra accuracy
The claim is not a sharper position. It is a position at all, where the alternative is dead reckoning and hope.
Algorithm and hardware
We are building the algorithm and a hardware stack that feed position straight into the vehicle control system.
The mission continues
The vehicle stays on route to a known site instead of turning back at the moment the ground stops being visible.
Why events
The event camera takes over when the vision camera cannot see.
A vision camera takes pictures. An event camera only notices change, so it still works when the picture would be too dark, too bright, or too fast.
| How they differ | Vision camera | Event camera |
|---|---|---|
| How it sees | Vision camera Takes complete photos at a fixed pace, like a video camera | Event camera Every pixel reports the instant the light in front of it changes |
| Fast motion | Vision camera Fast movement blurs the photo into a smear | Event camera Stays sharp even at high speed |
| Harsh light | Vision camera Bright parts wash out or dark parts turn black | Event camera Sees detail in sun and shadow at the same time |
| Low light | Vision camera Needs long exposures, which add blur and noise | Event camera Still picks out edges in moonlight |
| Glare | Vision camera Bright light spills over and hides the ground | Event camera The ground stays visible next to a bright sky |
| Light span | Vision camera ~60 dB Copes with a thousand-to-one span between darkest and brightest | Event camera 120 dB Copes with a million-to-one span, full sun to deep shadow |
| Update rate | Vision camera 30 / second One photo roughly every 33 milliseconds | Event camera 10,000 / second Continuous stream of pixel updates |
| Reaction time | Vision camera ~33 ms A change only shows up in the next photo | Event camera <1 ms A change is reported the moment it happens |
| Power and data | Vision camera Records the full picture all the time, even when nothing moves | Event camera Pixels that see no change stay silent and use nothing |
Team
Built across the whole stack.
Two co-founders. A computer-vision researcher and an aerospace engineer. One writes the event-camera mathematics. The other has shipped flight software and flown the machines.
Contact
Let us show you the way.
Part of the Slam Dunk Labs community.
Get in touch