Search papers, labs, and topics across Lattice.
The paper introduces Think-in-Control Vision-Language-Action (TIC-VLA), a novel framework that explicitly models and compensates for the inherent latency between semantic reasoning and real-time action generation in robot navigation tasks. TIC-VLA conditions action generation on delayed vision-language semantic states and explicit latency metadata, enabling the policy to anticipate and correct for asynchronous reasoning. The authors also propose a latency-consistent training pipeline and a new dynamic navigation simulation environment, DynaNav, to facilitate realistic evaluation.
Robots can now handle multi-second reasoning delays in dynamic environments thanks to a new architecture that explicitly models latency between perception and action.
Robots in dynamic, human-centric environments must follow language instructions while maintaining real-time reactive control. Vision-language-action (VLA) models offer a promising framework, but they assume temporally aligned reasoning and control, despite semantic inference being inherently delayed relative to real-time action. We introduce Think-in-Control (TIC)-VLA, a latency-aware framework that explicitly models delayed semantic reasoning during action generation. TIC-VLA defines a delayed semantic-control interface that conditions action generation on delayed vision-language semantic states and explicit latency metadata, in addition to current observations, enabling policies to compensate for asynchronous reasoning. We further propose a latency-consistent training pipeline that injects reasoning inference delays during imitation learning and online reinforcement learning, aligning training with asynchronous deployment. To support realistic evaluation, we present DynaNav, a physics-accurate, photo-realistic simulation suite for language-guided navigation in dynamic environments. Extensive experiments in simulation and on a real robot show that TIC-VLA consistently outperforms prior VLA models while maintaining robust real-time control under multi-second reasoning latency. Project website: https://ucla-mobility.github.io/TIC-VLA/
