Search
Fibocom AI Research Institute Attains an Average 2.6-Fold Acceleration in VLA Edge Inference
10 hour ago /
Read about 0 minute
Author：小编
On June 29, Fibocom announced that its AI Research Institute has achieved remarkable advancements in the realm of embodied AI. Leveraging its proprietary FiboVLA framework and edge inference optimization technology, the team successfully accelerated inference by an average factor of 2.6 across several mainstream VLA models. Moreover, they effectively deployed NVIDIA GR00T N1.5 onto a high-performance edge computing platform. The relevant accomplishments have been rigorously validated through both the LIBERO simulation benchmark dataset and a real-world desktop dual-arm robot environment, thereby offering robust engineering support for the efficient functioning of embodied AI models on robotic edge devices.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
