ForeTac: Predictive Contact Guidance for Generative Robot Policies

Visually similar actions can produce contact loss, jamming, slippage, or irreversible damage, while tactile feedback may arrive after a critical transition. Contact-rich manipulation therefore requires generative policies to anticipate and evaluate whether candidate actions will produce desirable contact outcomes, then adjust action generation before execution.

摘自论文摘要，摘要生成中。