EDGE-AI FUSION
AS-X1 fusion node — on-sensor inference at under 4 W per node.
The fusion node is the brain of the stack. It co-locates with the sensor head, ingests the raw modality streams over TSN, and emits a single perception frame (object list, occupancy grid, or semantic segmentation) to your vehicle or platform controller. No GPU server. No cloud round-trip.
ParameterAS-X1-S (single head)AS-X1-M (multi-head)
Ingest2 × 1 GbE TSN4 × 1 GbE TSN + 2 × CAN-FD
On-sensor inferenceUp to 8 TOPS (INT8)Up to 20 TOPS (INT8)
Fusion latency (end-to-end)≤ 18 ms≤ 24 ms
Power @ full load3.4 W5.8 W
Form factor60 × 45 mm, M.2 2230110 × 75 mm, custom carrier
SafetyASIL-B capableASIL-B / ASIL-D (dual-channel)
Compat.Compatible with NVIDIA Jetson Orin & Intel OpenVINO toolchainsSame
ROS 2
AUTOSAR 4.4 / R20-11
TSN (IEEE 802.1Qbv)
PTP grandmaster / slave
ONNX runtime
TensorRT (compatible)
Reference deployments
- 2024 Cross-modal sensor-fusion latency reduced by 68% across 240+ deployed units.
- 2023 ASIL-B perception pipeline for an autonomous yard tractor — production release.