Constraint-Aware Intent Estimation for Dynamic Human-Robot Object Co-Manipulation

TL;DR

This paper presents a framework that combines online estimation and control to facilitate robots in interpreting human intentions, and dynamically adjust their actions to assist in dynamic object co-manipulation tasks while considering both robot and human constraints.

Abstract

Constraint-aware estimation of human intent is essential for robots to physically collaborate and interact with humans. Further, to achieve fluid collaboration in dynamic tasks intent estimation should be achieved in real-time. In this paper, we present a framework that combines online estimation and control to facilitate robots in interpreting human intentions, and dynamically adjust their actions to assist in dynamic object co-manipulation tasks while considering both robot and human constraints. Central to our approach is the adoption of a Dynamic Systems (DS) model to represent human intent. Such a low-dimensional parameterized model, along with human manipulability and robot kinematic constraints, enables us to predict intent using a particle filter solely based on past motion data and tracking errors. For safe assistive control, we propose a variable impedance controller that adapts the robot's impedance to offer assistance based on the intent estimation confidence from the DS particle filter. We validate our framework on a challenging real-world human-robot co-manipulation task and present promising results over baselines. Our framework represents a significant step forward in physical human-robot collaboration (pHRC), ensuring that robot cooperative interactions with humans are both feasible and effective.

Figures (10)

• Figure 1: Our method uses particle filters to predict full 6 DoF intent and a variable impedance control scheme to assist the human, while being aware of the robot's kinematic constraints. This is achieved without any external force-torque (F/T) sensing. Schematic representation of robot state, $\mathbf{x}_r$, and force applied by human and robot in a co-carrying task. $\mathbf{u_\text{ext}}$ and $\mathbf{u}_h$ represent the wrenches applied to the object $\mathbf{x}$ by the robot and the human, respectively. The red, green, and blue arrows in the inertial frame Origin correspond to the x-axis, y-axis, and z-axis.
• Figure 2: Control diagram of our proposed approach. We estimate dynamics matrices $\widehat{\mathbf{A}_p}, \widehat{\mathbf{A}_o}$ and attractors $\widehat{\mathbf{p}^*}, \widehat{\mathbf{q}^*}$ online, these estimates along with their confidence $c_p, c_o$ are sent to the DS Variable Impedance Controller to generate desired force $\textcolor{orange}{\mathbf{u}_r^d}$. Lastly, a torque $\boldsymbol{\tau}_\theta$ satisfying the constraints is computed to send to the robot. Eq. \ref{['eq:combined_dyn_simpl']} refers to the combined dynamics in the Cartesian space.
• Figure 3: A demonstration of the human and robot co-carrying. Motion Capture system is employed to detect human kinematics. The manipulability ellipsoids of both human hands are depicted in blue. The ellipsoid associated with the hand closer to the pot is utilized to locally reshape the particles' noise, thereby influencing the goal distribution. Particles are visualized as white dots
• Figure 4: This demonstration showcase the proposed method in 6D. The operator has to move the content in the steel pot into another pot, with an obstacle in the way. In (a) and (b), the estimated $x,y$ and yaw goal changes as a result of external input in these dimensions. In (c) and (d), the estimated $z$ and roll goal changes as the operator pushes down and rotate the pot.
• Figure 5: Box plots of task completion time (seconds) over all trials, for the proposed method and baselines. The metric is summarized by the min and max values (black lines), first and third quartiles (box), median (orange line), and outliers (circle).

Theorems & Definitions (3)

• Theorem 5
• Lemma 6
• Lemma 7