The 6D pose estimation for metal parts is essential in industrial robotic applications. Although the 6D pose estimation methods that rely on object-specific training have gained extensive concern, these methods can’t generalize to novel objects. Recent novel object pose estimation methods are solving this issue using task-specific fine-tuned CNNs for deep template matching. However, these methods require expensive training procedures and don’t consider pose distribution of rendered templates. Recently, foundation models show strong representation learning ability, and can encode both the high spatial information as well as semantic information. In this study, we present a metal parts’ zero-shot 6D pose estimation method by foundation model without re-training, and incorporate prior information of metal parts’ poses to generate rendered templates that align with the pose distribution in real world. DINOv2, a recent vision foundation model with impressive generalization capabilities, is employed for matching rendered templates against query images of metal parts. Additionally, we introduce the optimal transport as a similarity metric. Then, the Gluestick is utilized to establish local keypoint correspondences, which enable deriving geometric correspondences and are used for estimating the metal part’s 6D pose with PnP/RANSAC. Given an industrial scenario, we first estimate metal parts’ pose, select and save high-confidence pose results to pose buffer, which is used to update the templates. The experiments on MP6D and ROBI datasets showcase that the proposed method has better performance than MegaPose. We also conduct real world experiments, which demonstrate the robustness of the proposed method. Code is available at https://github.com/sunhan1997/IndusPose.

To address challenges associated with high-dimensional coordination and strict physical constraints in fixed-base space manipulator tool transfer under zero gravity, this paper proposes a preference-conditioned safe reinforcement learning algorithm, which is called $\alpha$PPO-LAG ($\alpha$-Preference Proximal Policy Optimization with Lagrangian). The algorithm introduces a preference factor $\alpha$ to regulate the trade-off between task performance and physical safety through constraint-aware policy optimization. Meanwhile, an adaptive Lagrangian regulation mechanism based on constraint estimation is developed to improve safety satisfaction during training. Experimental results demonstrate that with $\alpha =0.5$, the proposed method achieves an average reward of 990, outperforming PPO-LAG with 881 and CPO with 915. Furthermore, $\alpha$PPO-LAG obtains a safety score of 0.935 while reducing the contact-force violation rate to 1.0% and maintaining a low end-effector velocity violation rate of 10.0%. The empirical safety-performance analysis reveals effective operating points under different safety preferences, providing a practical solution for safe and compliant fixed-base tool transfer in zero-gravity environments.