Virtual Physical Coupling of Two Lower-Limb Exoskeletons

Emek Baris Küçüktabak, Yue Wen, Matthew R. Short, Efe Demirbas, Kevin M. Lynch, José Luis Pons

Introduction

In the field of robotic rehabilitation, the physical interaction between a therapist and a patient is a critical component of motor learning. However, translating this "dyadic" interaction—where two people coordinate movements through touch—into robotic systems has historically been limited to simple upper-limb tasks or stationary lower-limb exercises. This paper, presented at ICORR 2023, introduces a novel framework for haptically linking the legs of two individuals during ambulatory walking. By creating "virtual" mechanical connections between two lower-limb exoskeletons, the researchers provide a new tool for exploring human-human coordination and developing more intuitive robotic gait assistance.

The Challenge of Gait Interaction

Most research into physical human-robot or human-human interaction focuses on single-joint movements, such as moving a joystick or rotating an ankle while seated. Walking is significantly more complex, requiring multi-joint coordination across the hip and knee, as well as the maintenance of dynamic balance. Prior to this work, there was a lack of infrastructure capable of simulating physical links between two people during full-body walking tasks. Such a system is a prerequisite for understanding how a robotic "partner" might best assist a patient in a way that mimics a human therapist’s guidance.

System Architecture and Methodology

The researchers utilized two ExoMotus-X2 lower-limb exoskeletons, which provide active torque at the hip and knee joints. These devices were integrated through a custom software architecture consisting of two primary layers:

  1. Virtual Coupling Node: This high-level controller acts as the system's "brain." It monitors the real-time positions and velocities of both users and calculates the theoretical forces that would exist if the users were physically connected by mechanical springs and dampers.
  2. Whole-Exoskeleton Closed-loop Compensation (WECC): This low-level controller ensures the exoskeleton motors accurately deliver the calculated interaction torques. A key challenge addressed by the WECC is compensating for the weight and internal friction of the heavy exoskeleton hardware, ensuring the users feel the "virtual link" rather than the robot itself.

Interaction Modalities

The framework supports several modes of interaction to test different coordination strategies:

  • Joint Space Coupling: Virtual torsional springs are applied directly to the hip and knee joints to synchronize joint angles.
  • Task Space Coupling: Virtual linear springs are placed between the users' ankles in 2D space, focusing on the trajectory of the foot.
  • Directionality: The system can operate bidirectionally (both users feel each other) or unidirectionally, where a "leader" guides a "follower" without feeling the follower's resistance.

Experimental Validation

The system was validated through a preliminary study involving a healthy dyad (a pair of able-bodied users). The experiment recorded joint angles, interaction torques, and gait timing across three conditions: no connection (NC), soft stiffness, and hard stiffness.

Key Insights

  • Physics as a Communication Channel: The virtual coupling effectively functions as a haptic language. Even in the absence of visual cues, the "pull" of the virtual spring allowed users to synchronize their gait timing and step length intuitively.
  • Stiffness and Agreement: The study found a direct correlation between the stiffness of the virtual connection and the level of movement synchronization. In the "hard" connection mode, the difference between the partners' joint angles was reduced by more than 50% compared to independent walking.
  • Haptic Guidance: In unidirectional trials, a leader was tasked with walking in an exaggerated, asymmetric pattern. The follower, despite having no visual information about the leader’s movements, was successfully guided into a matching asymmetric gait through the virtual forces alone.

Results

The quantitative results demonstrated the system's efficacy in enforcing coordination:

  • Synchronization Improvement: In the "No Connection" phase, users exhibited a peak hip angle difference of approximately 20°. Under a "Hard" virtual connection, this difference was suppressed to approximately 8°.
  • Error Reduction: The mean absolute difference in joint angles across the entire gait cycle was reduced from ~10–11° (NC) to just ~3° during the hard connection trials.
  • Task Space Performance: During stationary tests, the system demonstrated the ability to drag a follower’s ankle vertically by 30 cm in response to a 40 cm movement by the leader, using a 250 N/m virtual spring.

Limitations and Future Outlook

While the study proves the feasibility of virtual physical coupling, several hurdles remain. The current experiments were conducted at a slow speed (0.8 km/h) to ensure safety, as the system does not yet include a dedicated balance controller to prevent one user from inadvertently tripping the other. Additionally, the preliminary results are based on a single dyad; future research will involve larger, more diverse groups to account for variations in body type and motor skill.

The ultimate goal of this technology is clinical application. By coupling a stroke survivor to a healthy gait template—provided either by a therapist or a pre-recorded model—researchers hope to determine if this haptic "partnership" can accelerate motor recovery more effectively than current, non-interactive robotic therapies.

Summary updated

BibTeX
@inproceedings{kucuktabak2023virtual,
  author = {Emek Baris Küçüktabak and Yue Wen and Matthew R. Short and Efe Demirbas and Kevin M. Lynch and José Luis Pons},
  title = {{Virtual Physical Coupling of Two Lower-Limb Exoskeletons}},
  booktitle = {ICORR},
  year = {2023},
  doi = {10.1109/icorr58425.2023.10304601},
}
Figure