Search papers, labs, and topics across Lattice.
This paper introduces TacForeSight, a novel force-conditioned tactile foresight framework designed for real-time contact-rich manipulation in robotics. By utilizing a tactile world model (TacForceWM) that predicts tactile dynamics from dual-finger observations and high-frequency force signals, the system effectively anticipates contact changes and adapts its control strategies. Experimental results demonstrate that TacForeSight significantly outperforms existing methods in handling dynamic contact disturbances across various tasks, showcasing its potential for enhancing robotic manipulation capabilities.
TacForeSight outperforms traditional methods by enabling robots to anticipate and adapt to dynamic contact changes in real-time manipulation tasks.
Contact-rich manipulation requires robots to continuously perceive and regulate evolving physical interactions under dynamic contact transitions or complex surface geometries. Recent imitation learning methods improve contact-aware control by incorporating tactile or force feedback, but they rarely model the asymmetric spatiotemporal roles of global force and local tactile sensing. To address this, we propose TacForeSight, a lightweight force-conditioned tactile foresight framework for real-time manipulation. The core component is TacForceWM, a tactile world model that predicts short-horizon tactile latent dynamics from dual-finger tactile observations conditioned on high-frequency wrist force and torque signals. Another key component, the Predictive Tactile-Conditioned Policy, leverages the predicted latents as anticipatory contact priors, models the current-to-future tactile evolution via cross-attention, and adaptively fuses visuo-tactile features through a tactile-guided gating module. By forecasting purely within a compact latent space, TacForeSight enables proactive contact reasoning with efficient real-time inference suitable for high-frequency manipulation control. Real-robot experiments on five representative tasks and three in-process perturbation settings show that TacForeSight consistently outperforms existing baselines, particularly under dynamic contact disturbances. All models and datasets will be made publicly available on the project website at https://tacforesight.github.io/ProjectPage.
NUS