LATIDIA · Robótica

Más allá del servicio de LLM: Caracterización de las cargas de trabajo de visión-lenguaje-acción para el diseño de sistemas de IA incorporados

arXiv: 2610.05062v1Announce Type: cross Resumen: Los modelos de visión-lenguaje-acción (VLA) traducen las observaciones multimodales en acciones robóticas de bajo nivel. Durante el funcionamiento del robot, cada periodo de control establece un plazo de inferencia

La noticia

arXiv:2610.05062v1 Announce Type: cross Abstract: Vision-language-action (VLA) models translate multimodal observations into low-level robot actions. During robot operation, each control period sets an inference deadline, and overruns leave the robot acting on stale observations, reducing task success. Meeting this deadline motivates on-device or nearby edge execution, where a single robot requires batch-1 inference outside the design point of LLM serving systems. Although VLA architectures combine familiar vision-language, autoregressive, and diffusion-style components, their runtime behavior in this batch-1 control setting remains uncharacterized. We characterize four representative VLA models on an edge GPU server and two onboard SoCs, using single-inference profiling and 43,200 closed-loop episodes. Action tensor dimensionality determines whether a stage