TacMan-Turbo: Proactive Tactile Control for Robust and Efficient Articulated Object Manipulation

TL;DR

TacMan-Turbo is introduced, a novel proactive tactile control framework for articulated object manipulation that mitigates this fundamental trade-off between effectiveness and efficiency and demonstrates that the long-standing trade-off between effectiveness and efficiency in articulated object manipulation can be successfully resolved without relying on prior kinematic knowledge.

Abstract

Adept manipulation of articulated objects is essential for robots to operate successfully in human environments. Such manipulation requires both effectiveness--reliable operation despite uncertain object structures--and efficiency--swift execution with minimal redundant steps and smooth actions. Existing approaches struggle to achieve both objectives simultaneously: methods relying on predefined kinematic models lack effectiveness when encountering structural variations, while tactile-informed approaches achieve robust manipulation without kinematic priors but compromise efficiency through reactive, step-by-step exploration-compensation cycles. This paper introduces TacMan-Turbo, a novel proactive tactile control framework for articulated object manipulation that mitigates this fundamental trade-off. Unlike previous approaches that treat tactile contact deviations merely as error signals requiring compensation, our method interprets these deviations as rich sources of local kinematic information. This new perspective enables our controller to predict optimal future interactions and make proactive adjustments, significantly enhancing manipulation efficiency. In comprehensive evaluations across 200 diverse simulated articulated objects and real-world experiments, our approach maintains a 100% success rate while significantly outperforming the previous tactile-informed method in time efficiency, action efficiency, and trajectory smoothness (all p-values < 0.0001). These results demonstrate that the long-standing trade-off between effectiveness and efficiency in articulated object manipulation can be successfully resolved without relying on prior kinematic knowledge.

Figures (14)

• Figure 1: Schematic overview of proactive tactile proactive control framework, TacMan-Turbo. Our framework integrates three sequential components that enable proactive manipulation: (a) In-hand pose estimation extracts tactile contact patterns from gripper sensors to determine the relative transformation between current gripper pose $T_{g_i}$ and handle pose $T_{h_i}$. (b) Handle pose prediction analyzes sequential pose estimates ($T_{h_1}$, $T_{h_2}$) to extract local kinematic patterns, enabling prediction of the future handle pose $\tilde{T}_{h_3}$. (c) Offset velocity computation generates the optimal control signal by calculating an offset velocity $\boldsymbol{u}_i^{g_i}$ that, when combined with base velocity $\boldsymbol{u}_0^{g_i}$, creates a resultant motion $\boldsymbol{u}_0^{g_i}+\boldsymbol{u}_i^{g_i}$ that aligns future gripper position with predicted handle motion along path $J$. By proactively adjusting for potential contact deviations, this approach ensures smooth, continuous manipulation from start point $J_s$ to end point $J_e$ without requiring separate correction phases.
• Figure 2: Simulation test environment and object categories. Our evaluation encompasses three progressively complex articulated object categories: (a) 50 prismatic-joint objects featuring linear motion paths, representing the most kinematically straightforward mechanisms; (b) 50 revolute-joint objects with circular motion paths, introducing rotational complexity; and (c) 100 complex articulated objects with trajectories derived from Bézier curves of orders 2--5, simulating sophisticated real-world mechanisms with variable curvature. This systematically designed test suite enables rigorous performance assessment across a spectrum of kinematic complexity. Each case is standardized using initial robot configurations, ensuring fair comparison across manipulation strategies.
• Figure 3: Qualitative results of simulation studies. This visualization demonstrates TacMan-Turbo's performance across three representative cases from our object categories: prismatic (left), revolute (middle), and complex articulation (right). The top row displays trajectory tracking performance with actual handle positions (blue), predicted handle positions (orange), gripper positions (green), and contact points (red), while the bottom row shows corresponding tactile feedback patterns. For objects with consistent kinematic patterns (prismatic and revolute), TacMan-Turbo's prediction mechanism generates remarkably accurate estimates with minimal residual error, enabling precise trajectory alignment. When confronting complex articulations with sudden directional changes (right column), prediction accuracy temporarily decreases at inflection points, but the system rapidly converges back to the optimal path through iterative correction. The magnified insets highlight these prediction characteristics, demonstrating how TacMan-Turbo maintains effective manipulation even when faced with kinematically challenging scenarios. Contact deviation vectors (orange arrows, magnified 5× for visibility) further illustrate how the system continuously adjusts to maintain optimal contact. More qualitative results are available in the \suppUrlSI.
• Figure 4: Prediction errors of various predictive models.The prediction error distributions for three kinematic models---constant velocity, constant acceleration, and constant jerk---are presented and compared. All methods demonstrate comparable performance overall. However, higher-order models exhibit superior accuracy under low-noise conditions, as they better capture the underlying motion dynamics. Conversely, these same models show degraded performance in high-noise environments due to the cumulative effects of numerical differentiation, which amplifies measurement noise with each successive derivative computation. Box plots display time distributions with medians (center lines), interquartile ranges (boxes), and distribution extent (whiskers at $1.5\times$IQR). BC-n represents Bézier curves of order n.
• Figure 5: Time efficiency comparison (logarithmic scale, seconds). Evaluation of task completion times reveals TacMan-Turbo's superior efficiency compared to Tac-Man across all tested articulation types. Our method consistently completes manipulation tasks in significantly less time, with performance advantages particularly pronounced for complex articulations. Statistical analysis confirms these improvements are highly significant (****$p<0.0001$) throughout all joint categories. BC-n denotes Bézier curves of order n; box plots follow conventions established in \ref{['fig:sim-results-prediction']}.