Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots

TL;DR

Embodied.cpp is a portable C++ inference runtime for embodied AI models, enabling efficient deployment across heterogeneous devices and robots. It captures a shared execution path and provides modular multi-rate execution and latency-first fused inference. This improves deployment efficiency while preserving high accuracy.

Why It Matters

Embodied.cpp addresses the fragmentation of AI model deployment, allowing for more widespread adoption of embodied AI. Its portability and efficiency enable the use of embodied AI in various applications, including robotics and edge devices. This can lead to significant advancements in fields like robotics and autonomous systems.

Key Findings

• successful closed-loop execution with high task success rates
• reduced memory usage in WAM benchmark
• improved deployment efficiency across diverse model architectures