Embodied AI Glossary中文

Horizon Robotics OpenExplorer

地平线天工开物Advanced

Horizon Robotics' toolchain for converting, quantizing, and deploying AI models on its own BPU chips.

OpenExplorer (天工开物) is an AI development platform and toolchain from Horizon Robotics, built to serve its chips that carry a BPU (Horizon's own neural-network processor). A trained PyTorch or ONNX model can't run directly on this kind of chip — it first needs to be quantized (converting floating-point weights to a lower-precision format like INT8), compiled into chip instructions, and then loaded through a runtime library — and OpenExplorer provides the tools for these steps, along with example models and performance-profiling tools. Embodied-AI developers deploying vision or policy models on D-Robotics' RDK-series development boards commonly end up using this toolchain.

ExampleExport a YOLO detection model to ONNX, post-training-quantize and compile it with OpenExplorer's tools, then deploy it to an RDK development board for real-time detection.

Also called
OpenExplorer
Related
Horizon Robotics · D-Robotics RDK S100 · Post-Training Quantization · Neural Processing Unit (NPU) · On-Device / Edge Deployment · Open Neural Network Exchange (ONNX)
Sources
地瓜机器人开发者社区 (Chinese)

See it in the full glossary →