Is this page helpful?

Sensor Input Processing Pipeline#

SIPP is the data generation unit of HOISA. It processes sensor inputs to identify safety-relevant conditions and reports them downstream to SEI. Behavior differs between Tegra-based and x86-64-based platforms.

Architecture#

SIPP has two logical parts:

Physical Sensors — Monitor environment parameters. Most common: cameras (image streams). Others: temperature, pressure, flow rate, smoke density.
Software Framework — Processes sensor data against defined safety conditions. May use Deep Learning or legacy algorithms.

Reference SIPP Implementation

The HOISA Quick Start Guide uses the VSS Warehouse Blueprint - 2D Vision AI (NVIDIA Metropolis). Reference capabilities: person, forklift, and AMR detection with ROI, tripwire, and proximity event generation.

Platform-Specific Hardware Acceleration

On Tegra, SIPP may use these hardware engines:

Engine	Use case
iGPU (Integrated Graphics Processing Unit)	DNN inference; best for parallelizable non-DNN
PVA (Programmable Vision Accelerator)	Non-DNN computer vision algorithms
DLA (Deep Learning Accelerator)	Accelerated DNN inference

On x86-64, discrete GPU (dGPU) typically drives SIPP; CPU-based implementations also exist.

Connecting SIPP to HOISA

The SIPP-to-HOISA interface is unified across platforms. Connection options:

Method	Description
SEI interface library	Link directly to `libnvpss_interface.so`
Message brokering	Publish to Kafka, Redis, or similar
PSSCom library	Use the PSSCom communication library