Core Viewpoint - The article discusses the importance of accurate human intent estimation and effective role allocation in physical human-robot collaboration (pHRC), highlighting the limitations of current methods and introducing a new framework called DTRT to address these challenges [1][5][14]. Group 1: Challenges in pHRC - Achieving seamless collaboration between robots and humans requires effective strategies for accurate human intent estimation and dynamic adjustment of robot behavior [5]. - Current intent estimation methods primarily rely on short-term motion data, which limits their ability to predict long-term changes in human intent [1][5]. - The inability to anticipate changes in human intent can lead to potential conflicts and negatively impact the safety and efficiency of collaborative tasks [1][5]. Group 2: Introduction of DTRT Framework - The DTRT framework, developed by researchers from the Chinese Academy of Sciences and CASBOT, utilizes a dual transformer-based approach to enhance human intent estimation and role allocation in pHRC [1][2]. - DTRT addresses the core issues of inaccurate intent estimation and inflexible role switching by employing a hierarchical structure that captures human intent changes through guided motion and force data [2][7]. Group 3: Key Features of DTRT - DTRT tightly integrates human intent estimation with role allocation, allowing for quick detection of intent changes and timely adjustments to reduce human-robot discrepancies [7][8]. - The framework's human intent estimation module processes both motion and force data, improving the accuracy of intent predictions [8]. - The role allocation mechanism based on differential cooperative game theory ensures that robot behavior aligns closely with human intent while maintaining the robot's autonomy [8][7]. Group 4: Experimental Validation - A series of comparative experiments were conducted to validate the effectiveness of the DTRT framework, demonstrating significant advantages in prediction accuracy and collaborative performance [9][10]. - Key performance indicators showed that under the DTRT framework, the average human-robot collaboration angle reached 76.4°, with a robot assistance level index of 1.5, and the system maintained a good collaboration state 61.8% of the time [10][13]. Group 5: Future Implications - The introduction of DTRT represents not only an algorithmic breakthrough but also an attempt to reconstruct human-robot relationships, providing a versatile and valuable technical pathway for the development of humanoid robots [14]. - The research approach and core mechanisms of DTRT are expected to be further expanded and deepened in various practical applications, including industrial manufacturing and complex operations [14].