Beetle Eyes

Universal platform for autonomous target tracking, recognition and hitting

Advantages

The automatic target recognition and tracking platform is designed to hit military vehicles and infantry in the environment, which has different electronic warfare equipment, in the cases when the control signals exchange and video transmission are not possible at the final phase of the flight (500-1000m distance to the target).

Universal solution

ㅤㅤㅤㅤ FPVs, aircraft type UAVs, loitering munitions, cruise missiles, guided bombs

Software protection

ㅤ - Reverse engineering protection implemented specifically for the case when the drone is captured by the adversary - Per-device license management

Precision

ㅤ Over 90% success rate in environments with electronic warfare equipment when the tracking starts at 500 meters distance

Easy to use

ㅤ Can be deployed by unexperienced UAV pilots by turning on the tracker after take off and adjusting the tracking point during the flight

Production at scale

ㅤ Reference hardware cost is as low as $200 per unit (software license excluded)

Compatibility

ㅤ with different flight controllers and autopilots: - F405, H743, others - Ardupilot, Betaflight, iNav

Portability

ㅤ Software can be ported to any hardware platform, which meets performance requirements

Carriers

The system can be mounted on various carriers:

- FPV drones 7", 10", and 13"
- Aircraft-type UAVs

- Loitering munitions similar to Russia’s “Lancet” drones

- Cruise missiles / guided bombs

The system can be mounted on various carriers:

- "Khrushch" aircraft type UAVs
- Variety of 7", 10"and 13" FPV drones

The system can be mounted on various carriers:

- Ardupilot
- Betaflight
- iNav

Specifications

Target capturing and tracking distance: 8mm lens / 50° FOV: up to 1000m

Target speed: depends on carrier drone capabilities. For a basic 10" drone, target speed can be up to 60km/h for both chase and head-on approaches.

Max crosswind: 12kt (6 m/s)

Tracker FPS: 30

Training dataset: proprietary dataset of 150,000+ labeled images

Tracker precision (on training dataset): 0.95

Detection models:

- military vehicles (tanks, personnel carriers, SAMs, trucks)

- military and civilian cars, vans, trucks

- flying targets

- infantry

Detector performance: 10 FPS

Image resolution: 640x480 (zoomed image area: 1440x1080)

Weight: <150g including camera hardware and lens

Gimbal support: 1, 2 axis

Pilot User
ㅤ Interface:

State:

Field testing videos

Click below to watch the videos

The system can be ported to a different hardware platform within 6-9 weeks.

Adaptation of the existing HW to a new UAV model: 3-4 weeks for setup and field testing

Hardware platform

Raspberry Pi 4b

RPi Global Shutter Camera

Coral USB stick

(for target detection)

target detector

For our target detector training, we've collected a dataset with 200K images (25+ hours of video) and labeled it with 12 different classes of targets (Person, BMP, Tank, Truck, ADS, MLRS, Artillery etc.).

The detector can be used to ensure the target hit in the case when for a substantial period of time the line-of-sight has been lost or the target is hidden in smoke.

Even without a detector, our computer vision algorithms and logic incorporated in the tracker module can still keep hold of the target in case of brief line-of-sight losses.

Currently we develop a fully autonomous scenario where the drone autonomously reaches a geo zone specified by the operator at the start of the mission, searches for the target according to the priority and hits it, with the assumption that there is no GPS and no connection between the drone and the operator.

Control and tracking algorithms are being tested in flight simulators (Uncrashed, XPlane, AirSim, PyBullet) because they allow:

- Specifying UAV parameters and performing initial PIDs tuning before field testing

- Algorithms testing for the scenarios when the targets are moving

- Algorithms testing in different wind conditions

Simulation

The platform development is being done in close cooperation with the 48th combat unit of AFU, who developed the aircraft type UAV "Khrushch", and with a large drone pilot school.