Detect people & large animals.
Heimdall’s mission is to warn you about approaching people and large animals, such as bears, while keeping false alarms to a minimum.
Optimized for three things, in this order
1Detection rate47 of 49 walks detected in the field
2False alarm rate1 false alarm on 6 field days
3Detection distanceTypical alarm at 38 m (126 ft)
Where the project is right now.
This is how far Heimdall has come towards its goals. I plan to have made solid progress on most of them by spring 2027.
Detect people
- Walking: Solid progress: Simulated, Field-tested. Simulated, by weather: 3 × 171 walks. Calm: 36 of 36. Wind: 90 of 90. Rain: 36 of 36. Wind and rain: 4–8 of 9. Detected: 166–170 of 171. Alarm at: 38–42 m. Field-tested, by weather: 6 field days. Calm: 32 of 33. Wind: 10 of 10. Rain: 5 of 5. Wind and rain: 0 of 1. Detected: 47 of 49. Alarm at: 38 m.
- Running: In progress: Simulated. Planned: Field-tested.
- In a group: In progress: Simulated.
Detect large animals
- Lone bear: In progress: Simulated.
- Bear with cubs: Planned: Simulated.
Ignore the rest
- Wind, rain and noise: Solid progress: Field-tested. Next up: Overnight. Field-tested, false alarms, by where they were counted: 6 field days. During walks: 1 in 49 walks. In the noise: 0 in 99 min. False alarms: 1.
- Small animals: Planned: Simulated.
How the results are counted
- Misses count. A walk that was not detected stays in every count, and counts as 0 m in a typical distance.
- False alarms count everywhere. During the walks, and in recordings of wind and background noise with nobody around.
- Simulations stay apart. A simulated result is a range over several runs, and is never added to a field result.
Three stages, each more real.
1simulation513 simulated walks
Simulated
Real footsteps from the field days, placed into real background recordings from different weather and ground conditions.
6field days47 of 49 walks detected
Field-tested
GPS-tracked walks towards the sensor in the field, so every detection and alarm has a measured distance.
0nightsThe first is next up
Overnight
The sensor left out overnight, just as it would be at a campsite, with cameras alongside to check that it detects and ignores what it should.
See every test in detail.
Every field day and simulation has its own entry in the Field Log, misses and false alarms included.
Open the Field Log