MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641119807 A) filed by Chennai Institute Of Technology on October 07, 2026, for A Heterogeneous Decoupled Edge-Computing Robotic Architecture For Embodied Ai.
Inventors include Sharveshe S S; Dr. Rajmal Joshi; Dr. Thanigaivel Raja. M; and Dr. Joly. M.
The application for the patent was published on October 09, 2026, under issue no. 41/2026.
Abstract: The integration of Large Language Models (LLMs) into physical robotic platforms represents a significant advancement toward Embodied Artificial General Intelligence. However, bridging the gap between high-latency cloud reasoning and low-latency hardware execution on edge devices remains a critical challenge. Project Aura introduces a heterogeneous, split-core architecture designed to create a highly responsive, multimodal Embodied AI agent using accessible hardware. The system employs a "Brain vs. Spine" topology: a Raspberry Pi-based Cognitive Cortex handles computationally intensive tasks, while a dedicated Arduino-based Reflex Core manages deterministic motor control. The cognitive layer utilizes highly optimized, multi-threaded sensor fusion. Environmental perception is achieved through an OpenCV Deep Neural Network (DNN) pipeline for real-time semantic object tracking, enabling autonomous "hunting" kinematics. Human-Robot Interaction (HRI) is driven by a custom offline wake-word engine (Vosk) featuring continuous raw-byte audio buffering to prevent ALSA stream collision, coupled with cloud-based natural language processing (GPT-4o-mini). To ensure contextual relevance, the agent utilizes a live Retrieval-Augmented Generation (RAG) pipeline via integrated web scraping. Furthermore, affective feedback is provided through a dynamic, hardware-accelerated "Liquid Face" UI that visually reflects the robot's internal processing states. By successfully decoupling complex multimodal cognition from real-time physical execution, Project Aura demonstrates a stable, scalable, and cost- effective blueprint for the next generation of interactive, autonomous robotic platforms.
Disclaimer: Curated by HT Syndication.
