A Cooperation Control Framework Based on Admittance Control and Time-varying Passive Velocity Field Control for Human-Robot Co-carrying Tasks

TL;DR

This work tackles safe, energy-aware human–robot co-carrying by coupling a deep-LSTM-based reference generator with admittance correction to proactively align robot motion to human intent, with a low-level, energy-compensation time-varying PVFC that enforces passivity and finite-time energy convergence. A fictitious flywheel augments the robot dynamics to enable a strictly passive closed-loop system, while a fractional-energy term and skew-symmetric port-Hamiltonian structure regulate power flow and stability. Theoretical proofs (Lyapunov-based) establish passivity, convergence of kinetic energy, and stable tracking; experiments with 18 participants demonstrate reduced interaction forces, lower power-flow variation, and improved task performance and workload metrics (p < $0.05$) versus baselines. The results indicate the framework can provide proactive assistance without sacrificing safety, offering a practical pathway to more capable and user-friendly human–robot co-manipulation in industrial and daily settings.

Abstract

Human-robot co-carrying tasks reveal their potential in both industrial and everyday applications by leveraging the strengths of both parties. Effective control of robots in these tasks requires managing the energy level in the closed-loop systems to prevent potential dangers while also minimizing motion errors to complete the shared tasks. The collaborative tasks pose numerous challenges due to varied human intentions in adapting to workspace characteristics, leading to human-robot conflicts. In this paper, we develop a cooperation control framework for human-robot co-carrying tasks constructed by utilizing reference generator and low-level controller to aim to achieve safe interaction and synchronized human-robot movement. Firstly, the human motion predictions are corrected in the event of prediction errors based on the conflicts measured by the interaction forces through admittance control, thereby mitigating conflict levels. Low-level controller using an energy-compensation passive velocity field control approach allows encoding the corrected motion to produce control torques for the robot. In this manner, the closed-loop robotic system is passive when the energy level exceeds the predetermined threshold, and otherwise. Furthermore, the proposed control approach ensures that the system's kinetic energy is compensated within a finite time interval. The passivity, stability, convergence rate of energy, and power flow regulation are analyzed from theoretical viewpoints. Human-in-the-loop experiments involving 18 participants have demonstrated that the proposed method significantly enhances task performance and reduces human workload, as evidenced by both objective metrics and subjective evaluations, with improvements confirmed by statistical tests (p < 0.05) relative to baseline methods.

Figures (13)

• Figure 1: Human--robot co-carrying task, in which predicted human intention for the robot may occasionally contain errors or perturbations, thereby causing human--robot conflicts.
• Figure 2: Proposed cooperation control framework for human--robot co-carrying tasks, including reference generator and low-level controller. For the reference generator, a deep LSTM model is trained to predict human motion $\left( \mathbf{\hat{x}},\mathbf{\dot{\hat{x}}} \right)$ according to the previously measured motion data $\left( \mathbf{\Phi } \right)$. This predicted motion is corrected in the event of prediction errors thanks to admittance control with external forces $\left( {{\mathbf{f}}_{ext}}\right)$ which are measured by a force sensor. According to the output of the reference generator $\left( {{{\mathbf{\dot{x}}}}_{a}}\right)$ and robot's states $\left( \mathbf{q},\,\,\mathbf{\dot{q}} \right)$, the reference trajectory $\left( \mathbf{Q},\,\,\mathbf{\dot{Q}} \right)$ is encoded by a time-varying velocity field $\left( {{\mathbf{V}}^{a}}\left( {{\mathbf{q}}} \right),t \right)$ of the low-level controller. Thereupon, control torques for the robot's actuators $\left( \pmb{\tau } \right)$ are produced based on improved PVFC.
• Figure 3: Deep LSTM architecture for motion prediction. The input layer is established by the interaction motion of the collaborative co-carrying tasks, encompassing $m$ vectors $\mathbf{\Phi }\left( t-\left( m-i \right)T_s \right)$, $i=1,2,\ldots,m$. The stacked LSTM layers are designed by sequentially stacking multiple LSTM networks, where the output of each preceding network serves as the input for the subsequent network. The output layer $\left(\mathbf{y}\left( t+T_s \right)\right)$ can calculate the future reference position and velocity for the robot partner.
• Figure 4: The experimental framework of the human--robot co-carrying task, including (a) the experimental setup where a rigid object is collaboratively grasped by a human and a robot; (b) the object at its initial position (O) intended to toward position (B); (c) the object at position (B) transported to one of the target positions (A, C, or D); and (d) the object at the parking location (D).
• Figure 5: Prediction performance of the deep LSTM model for each participant, (a) ML2E of position and (b) ML2E of velocity.

Theorems & Definitions (10)

• Definition 1
• Lemma 1
• Remark 1
• Theorem 1
• Theorem 2
• Theorem 3
• Corollary 1
• Lemma 2
• Lemma 3
• Remark 2