LightLasers and photonics
Vibration read from a distance for machine health. Gas leaks and plumes read by laser. Lidar for shape and range. Laser scanning for structural change. They fail in dust, smoke, rain and under a spoofed or dazzled return. We test lidar and laser readers under obscurants and attack, and we score plume and vibration AI against known truth.
RadioRadar
Sees through fog, dust, sleet and snow when cameras and lidar cannot. Finds voids and rebar behind concrete. Tracks drones near a facility. It fails on clutter, ghost echoes, interference and ice on the cover. We build sealed radar test sets and score how a system handles a jammed or spoofed return.
Sound and touchAcoustic and vibration
Machine health from sound and vibration. Leaks heard before they are seen. Sonar in murky water. Tap testing of structures. The failure is a model that learned the machine instead of the fault, and a robot that cannot hear past its own noise. Our test sets hold out whole machines, so a score means the fault was learned.
Heat and spectrumThermal and multispectral
Thermal for heat loss, hot spots and night work. Shortwave infrared and gated imaging through haze. Polarization against glare. Event cameras for fast motion. They fail on reflections, thermal crossover and a dirty window. We score imaging AI by condition with the true temperature or state known for every item.
PositionNavigation without GPS
Underground, indoors, under water, in a jammed area, off-world. Drift is expected. The hazard is a robot that drifts at full confidence. We score whether the system knows when its position can no longer be trusted, and how fast it says so.
Test dataSynthetic sensor test sets
Real base data with physics injected on top, or scenes rendered with known truth, so every item has an exact answer. Keys are sealed by hash before any run. Zero-signal items count false alarms. The synthetic set is checked against a real bench before it is used to score anyone.