{"id":67,"date":"2025-05-05T16:57:37","date_gmt":"2025-05-05T14:57:37","guid":{"rendered":"https:\/\/projects.dimes.unical.it\/music4d\/?page_id=67"},"modified":"2025-12-17T19:41:01","modified_gmt":"2025-12-17T18:41:01","slug":"lattrezzatura","status":"publish","type":"page","link":"https:\/\/projects.dimes.unical.it\/music4d\/lattrezzatura\/","title":{"rendered":"Equipment"},"content":{"rendered":"\n
\n\n

\n System Architecture Overview<\/strong>\n <\/h1>\n

\n High-Performance Edge Computing, Multi-Sensory Acquisition & Embodied AI\n <\/p>\n\n

\n\n
\n

\n The MUSIC4D<\/strong> technical architecture is designed to handle complex real-time processing of multimodal data. It combines a powerful Edge Node infrastructure for heavy computation, a delicate sensory interface for environmental inputs, and state-of-the-art robotics for physical interaction.\n <\/p>\n <\/section>\n\n

\n

\n \u26a1 Computational Core: Edge Nodes\n <\/h2>\n

\n At the heart of our processing capabilities are the Edge Nodes<\/strong>. These units are engineered to run sophisticated Deep Learning models (VLMs) and handle the “Tutti-bot” framework operations with low latency.\n <\/p>\n \n

\n

GPU Acceleration<\/h3>\n \n

\n To ensure maximum throughput for AI workloads, the nodes are equipped with high-performance graphics units. The architecture leverages the power of NVIDIA RTX PRO 6000 Workstations<\/strong> (or alternatively RTX A4500<\/strong>), providing the necessary CUDA cores for real-time inference, generative audio synthesis, and foundation model fine-tuning.\n <\/p>\n <\/div>\n

\n *Additionally, edge computing is supported by NVIDIA Jetson AGX Orin<\/strong> boards for localized robotic processing.\n <\/p>\n <\/section>\n\n

\n

\n \ud83c\udf99\ufe0f Sensory Interface: Audio Acquisition\n <\/h2>\n

\n To interpret the “voice” of the environment and the audience, the system utilizes a tiered microphone setup managed by a Focusrite Scarlett 18i20 (4th Gen)<\/strong> multi-channel converter.\n <\/p>\n\n

\n
\n

High-Fidelity Capture<\/h4>\n \n

\n For critical audio analysis requiring maximum detail, we employ the Schoeps Flexi Set CMC6 MK2<\/strong>. These microphones provide the transparency needed for delicate acoustic nuances.\n <\/p>\n <\/div>\n\n

\n

Ambient & Versatile Capture<\/h4>\n

\n To cover a broader range of inputs and ensure versatility across the performance area, the system integrates Behringer B5<\/strong> microphones for distributed audio sensing.\n <\/p>\n <\/div>\n <\/div>\n <\/section>\n\n

\n

\n \ud83d\udc41\ufe0f Visual Input\n <\/h2>\n

\n Complementing the audio, the visual perception system utilizes Sony FDR-AX43A<\/strong> (Night Vision) and Panasonic HC-VXF11EG<\/strong> 4K cameras. These feed high-resolution video streams into our VLM pipeline for emotion recognition and scene analysis.\n <\/p>\n <\/section>\n\n

\n

\n \ud83e\udd16 Embodied AI & Robotic Actuation\n <\/h2>\n

\n Transitioning from perception to physical interaction, the system integrates state-of-the-art humanoid robotics managed by generalist foundation models. This layer allows the “Tutti-bot” to inhabit the physical space alongside performers.\n <\/p>\n\n

\n \n
\n

Unitree G1 Humanoid<\/h3>\n \n

\n The physical avatar of the system is the Unitree G1<\/strong>. This humanoid agent offers high-torque joint movement and agility, serving as the kinetic output for the generated behaviors. It translates the AI’s “emotions” into gestures and spatial navigation.\n <\/p>\n <\/div>\n\n

\n

NVIDIA Project GR00T<\/h3>\n \n

\n Orchestrating the robot’s cognition is NVIDIA Project GR00T<\/strong>. This general-purpose foundation model enables multimodal understanding (language, video, and demonstration), allowing the robot to learn coordination and interact naturally rather than following hard-coded scripts.\n <\/p>\n <\/div>\n \n <\/div>\n <\/section>\n\n<\/div>\n","protected":false},"excerpt":{"rendered":"

System Architecture Overview High-Performance Edge Computing, Multi-Sensory Acquisition & Embodied AI The MUSIC4D technical architecture is designed to handle complex real-time processing of multimodal data. It combines a powerful Edge Node infrastructure for heavy computation, a delicate sensory interface for environmental inputs, and state-of-the-art robotics for physical interaction. \u26a1 Computational Core: Edge Nodes At the … <\/p>\n

Leggi tutto “Equipment”<\/span><\/a><\/p>\n","protected":false},"author":7,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":[],"_links":{"self":[{"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/pages\/67"}],"collection":[{"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/comments?post=67"}],"version-history":[{"count":24,"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/pages\/67\/revisions"}],"predecessor-version":[{"id":210,"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/pages\/67\/revisions\/210"}],"wp:attachment":[{"href":"https:\/\/projects.dimes.unical.it\/music4d\/wp-json\/wp\/v2\/media?parent=67"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}