PaperlayerView exact revision on arXiv

Source-linked AI summary

A Reconfigurable Bidirectional Cable-Driven Hip Exoskeleton with Swappable Bench/Backpack Dual-configuration Actuation

YuanLong Ji, Shuhan Xiang, Qihan Ye, Xingbang Yang

arXiv:2609.25639v1cs.ROcs.SYeess.SY

TL;DR

Bench and portable hip exoskeletons provide different resources, but separate prototypes limit continuity across laboratory and mobile scenarios. This paper presents a reconfigurable cable-driven platform with a shared wearable interface and swappable actuation. Human-worn trials validated switching between configurations and recorded walking data in both settings.

  • Problem

    Bench and portable exoskeletons offer different capabilities, yet commonly remain separate prototypes, limiting reuse of the same wearable structure and sensing arrangement across scenarios.

    Source: §I · p. 2

  • Method

    The platform combines a cable-free wearable hip interface with swappable bench- or backpack-mounted actuation, common end-effector sensing, and configuration-specific control paths.

    Source: §I · p. 2

  • Results

    30.1 ± 16.3 s overall switching time across 30 trials; experiments also validated bench-mounted motion tracking and recorded continuous walking data in both configurations.

    Source: A. Low-Level Actuation Response · p. 5, B. Actuation Switching and Comfort · p. 6, C. Human-Worn Data Recording · p. 6

  • Takeaways & Limitations

    The same wearable structure, quick-release interface, and sensing path can support laboratory testing, actuation switching, and mobile validation.

    Source: §VI · p. 6, §VII · p. 7

    Backpack comfort was lower, mainly associated with the larger power supply used for outdoor recording endurance; reducing carried power-system mass remains a next step.

    Source: §VI · p. 7

Abstract

from arXiv · show

Hip exoskeletons provide an important hardware basis for lower-limb rehabilitation and locomotor assistance. Laboratory rehabilitation assessment and system development require substantial actuation and computing resources, whereas mobile assistance requires untethered portability. Integrating both capabilities within one reusable platform remains a central design challenge. This paper presents a reconfigurable bidirectional cable-driven hip exoskeleton platform that rapidly switches between bench-mounted and backpack-mounted actuation while sharing one cable-free wearable hip interface. The platform modularly adapts the actuation configuration, end-effector sensing path, and low-level control interface. Each cable-driven end-effector weighs 0.405 kg, excluding the cable and actuation unit, and integrates an encoder and a torque sensor; experiments validated bench-mounted admittance-based motion tracking capability and backpack-mounted open-loop torque tracking. Human-worn experiments with three healthy participants used myoMOTION to evaluate the platform's wearable-side hip-motion sensing capability, verified bench-to-backpack and backpack-to-bench motion-ready switching across 30 trials in $30.1\pm16.3$ s, and formed a small-scale multimodal wearable-exoskeleton gait dataset for sensing validation and data-driven algorithm development, comprising 8 min bench-mounted treadmill records and 11 min backpack-mounted outdoor walking records. These results show that, by unifying the wearable structure, actuation interface, and sensing path, the proposed platform enables validation of the same hip exoskeleton in both bench-mounted and backpack-mounted configurations, providing reusable hardware for iterative development and applications across scenarios.

I. INTRODUCTION

Bench-mounted and portable hip exoskeletons offer different capabilities, but are commonly built as separate prototypes. This paper presents a shared wearable interface and sensing arrangement with swappable actuation for laboratory and mobile use.

I. INTRODUCTION

  • Bench-mounted systems support resource-rich, repeatable laboratory work, while portable systems enable outdoor use but must balance capability, endurance, sensing, computation, and safety against carried mass.Source: §I · p. 1
  • Separate bench and backpack prototypes limit reuse of the same wearable structure and sensing arrangement across development stages and application scenarios.Source: §I · p. 1
  • Cable and Bowden transmissions can place actuators away from the joint, but friction, hysteresis, pretension sensitivity, and changing cable configurations degrade actuator-to-end-effector consistency.Cross-scenario reuse therefore also requires compatible end-effector sensing, mechanical attachment, sensor acquisition, and actuator-control connections.Source: §I · pp. 1–2
  • The platform decouples actuation from a cable-free wearable interface and uses a common quick-release assembly for bench- or backpack-mounted units.Its end-effectors integrate torque and angle sensing, with wireless IMUs supporting comparable human-motion measurements after actuation replacement.Source: §I · p. 2

A. Task Requirements

The platform’s task requirements guide actuation selection for laboratory and portable assistance. Both configurations use the same wearable interface and bidirectional cable connection, with actuation units swapped at a common interface.

A. Task Requirements

  • Selected task torques range from 5.6–6.1 N m for 20% portable walking assistance to 19.9–24.0 N m for 40% sit-to-stand and ramp-ascent assistance.The corresponding component-level peak torques are 95.25 N m for the bench unit and 53 N m for the backpack unit.Source: A. Task Requirements · p. 2

B. System Overview

  • The bench unit provides off-board laboratory actuation, while the backpack unit carries actuation and lightweight control hardware on the waist-back side; both drive the same wearable hip interface.Source: B. System Overview · p. 2

C. Common Quick-Release/Assembly Interface and Cable Routing

  • A spring-loaded quick-release interface swaps end-effectors without changing the waist or thigh structure, while separate flexion and extension cable paths provide bidirectional transmission.Both configurations preserve the output direction, joint-angle convention, and end-effector torque measurement point.Source: C. Common Quick-Release/Assembly… · pp. 2–3

D. Bench-Mounted Cable-Driven Unit

The platform pairs distinct bench-mounted and backpack-mounted actuation hardware with a shared wearable hip interface and end-effector sensing arrangement.

D. Bench-Mounted Cable-Driven Unit

  • The bench-mounted unit places its motors, reducers, pretensioners, and stops on an external platform, with 95.25 N m peak actuator-side torque and approximately 360 deg/s output speed.External power and industrial computing support high-bandwidth acquisition and real-time algorithm debugging.Source: D. Bench-Mounted Cable-Driven Unit · p. 3

E. Backpack-Mounted Cable-Driven Unit

  • The backpack-mounted unit integrates AK10-9 V3.0 modules, pretensioners, and low-level control hardware in a waist-back housing, with 53 N m peak torque per module.A Jetson Nano, STM32 controller, and CAN bus support mobile recording and outdoor experiment extensions.Source: E. Backpack-Mounted Cable-Driven Unit · p. 3

F. Wearable-Side Lightweight Sensing Integration

  • Each quick-release/assembly end-effector weighs 0.405 kg, excluding the cable and actuation unit, and integrates a torque sensor and absolute encoder; shared type and wireless IMU layout preserve comparable measurements after replacement.The passive hip connection accommodates approximately ±50° internal-external rotation and ±40° abduction-adduction motion.Source: F. Wearable-Side Lightweight Sensing… · p. 3

III. SWITCHABLE CONTROL

The control architecture retains common sensing variables while adapting control paths and hardware to each actuation configuration. Bench-mounted operation uses end-effector torque in an admittance model to generate a position command.

A. Common Sensing and Electrical Interface

  • Hip angle, end-effector torque, and wireless IMU data use common sensing interfaces, while each configuration connects to its own computing and motor-control hardware.Bench-mounted control uses data acquisition and motion-control interfaces; backpack control uses an embedded controller and CAN bus.Source: A. Common Sensing and Electrical… · p. 3

B. Bench-Mounted Admittance-Based Motion Tracking

  • In bench-mounted admittance tracking, torque offset by a static baseline maps through virtual inertia, damping, and stiffness to a position displacement, then a position servo tracks the desired position.The desired position is the initial reference plus that displacement; the control architecture is shown in Fig. 3.Source: B. Bench-Mounted Admittance-Based… · pp. 3–4, Figure 3 · p. 4

C. Backpack-Mounted Torque Control

The backpack-mounted configuration uses open-loop torque control to produce bidirectional hip output, mapping a scaled reference torque to actuator input.

C. Backpack-Mounted Torque Control

  • The backpack-mounted unit uses open-loop torque control for bidirectional hip output.The sign of the target torque selects the flexion-side or extension-side cable.Source: C. Backpack-Mounted Torque Control · p. 4
  • The target torque is the reference curve scaled by γ, with its sign selecting the flexion- or extension-side cable.τd(t) = γτref(t), where γ is the magnitude scaling factor.Source: C. Backpack-Mounted Torque Control · p. 4
  • The controller converts target torque to actuator input using the backpack actuator’s nominal torque coefficient.The end-effector torque sensor independently measures output torque for evaluating tracking through the cable transmission.Source: C. Backpack-Mounted Torque Control · p. 4

IV. EXPERIMENTAL SETUP

The experiments assess low-level response in both actuation configurations and whether the shared wearable interface supports human-worn use. Bench-mounted testing uses admittance-based motion tracking, while backpack-mounted testing applies bidirectional open-loop torque commands.

IV. EXPERIMENTAL SETUP

  • Three healthy participants completed human-worn walking, switching, and comfort tests while end-effector signals, wireless IMUs, and myoMOTION hip kinematics were recorded synchronously.The experiment plan targeted low-impedance bench response, measurable backpack torque output, and continued use of the same interface across treadmill and outdoor walking.Source: §IV · p. 4

A. Low-Level Actuation Response Test

  • Bench-mounted human-worn testing used low-inertia, low-damping, zero-stiffness admittance settings to evaluate synchronized hip-angle tracking and interaction-torque measurement.Encoder-based angle was compared with myoMOTION reference motion while baseline-corrected end-effector torque drove the position-servo cable actuator.Source: A. Low-Level Actuation Response Test · p. 4
  • Backpack-mounted tests applied a 10 N m ramp-hold command and a 5 N m, 0.25 Hz bidirectional sinusoid to evaluate sustained and repeated loading.Five complete ramp-hold sequences and five sinusoidal windows, each containing ten cycles, were retained.Source: A. Low-Level Actuation Response Test · p. 4
  • Backpack torque response was summarized by ramp-up slope, plateau measured-to-commanded torque ratio, and sinusoidal fitted-amplitude ratio.The ramp-up slope used a zero-intercept fit over 2–9.5 N m; the plateau measure was averaged during the hold.Source: A. Low-Level Actuation Response Test · pp. 4–5

B. Actuation Switching and Comfort Test

The switching test measured how long it took to restore motion readiness after exchanging actuation configurations, alongside comfort ratings before and after walking.

B. Actuation Switching and Comfort Test

C. Human-Worn Data Collection

Human-worn records paired bench-mounted treadmill walking with backpack-mounted outdoor walking, and were organized to support sensing validation and data-driven gait and interaction algorithms.

C. Human-Worn Data Collection

  • Each participant completed an 8 min bench-mounted treadmill walk at 1.0 m/s and an outdoor backpack-mounted walk averaging 11.17±1.47 min.The outdoor route included level walking, ramps, stairs, turns, start-stop transitions, and self-selected speed changes.Source: C. Human-Worn Data Collection · p. 5
  • Encoder-based hip angles were aligned to myoMOTION references and evaluated using calibrated RMSE, MAE, and correlation coefficient.Channel acquisition quality was summarized with valid-frame ratios and effective update rates.Source: C. Human-Worn Data Collection · p. 5
  • The synchronized records were organized as a small-scale multimodal gait dataset with anonymized raw and aligned data, a data dictionary, processing scripts, and an anonymized outdoor-route description.The paired inputs support feasibility assessment of wearable sensing and development of end-to-end data-driven gait and interaction algorithms.Source: C. Human-Worn Data Collection · p. 5

V. EXPERIMENTAL RESULTS

The two actuation configurations produced measurable low-level responses, and switching between them restored a motion-ready system without changing the wearable hip interface. Bench-mounted tracking followed the myoMOTION reference with low interaction torque, while backpack-mounted responses were repeatable under the tested inputs.

A. Low-Level Actuation Response

  • 1.48 N m mean absolute baseline-corrected interaction torque accompanied bench-mounted tracking, with encoder hip angles matching the myoMOTION pattern at 1.80◦ RMSE and 0.994 correlation.Fig. 6(a,b) presents the interaction torque and time-aligned angle traces.Source: A. Low-Level Actuation Response · p. 5, Figure 6 · p. 6
  • A 10 N m ramp-hold and a ±5 N m, 0.25 Hz sinusoidal input produced repeatable backpack-mounted end-effector responses.The ramp holding torque was 5.84±0.20 N m; the sinusoidal response had a 2.74±0.13 N m fitted amplitude and 10.4 ± 0.7◦ phase lag.Source: A. Low-Level Actuation Response · p. 6, Figure 6 · p. 6

B. Actuation Switching and Comfort

  • 30.1 ± 16.3 s overall switching time restored a motion-ready system without changing the wearable hip interface; comfort fell more after backpack-mounted outdoor walking.Switching time had a 23.8 s median; comfort decreased from 6.3 to 3.7 after outdoor walking, versus 6.7 to 5.3 after bench-mounted treadmill walking.Source: B. Actuation Switching and Comfort · p. 6

C. Human-Worn Data Recording

The human-worn records assess hip-motion sensing and interaction during walking in both actuation configurations. The discussion identifies control, comfort, and validation work that remains.

C. Human-Worn Data Recording

  • Fig. 7 reports motion-readiness switching times and participant comfort scores before and immediately after walking in each configuration.Switching plots show individual trials with mean ± SD error bars; comfort scores range from 1 to 7, with higher scores indicating better comfort.Source: Figure 7 · p. 6
  • End-effector encoder and torque channels provided stable high-rate records, while wireless IMU updates were more variable; the recordings support sensing evaluation and synchronized-signal model development.Measured absolute end-effector torque was 1.05 ± 0.32 N m during bench-mounted admittance tracking and 0.68 ± 0.35 N m during backpack-mounted unpowered walking.Source: C. Human-Worn Data Recording · p. 6, Figure 8 · p. 7

VI. DISCUSSION

  • The listed platform combines end-effector torque sensing, encoder-based joint measurement, wireless IMUs, and bench- and backpack-mounted operation in one shared sensing path.Bench mounting offers stable power, sensing expansion, high-bandwidth acquisition, and computation that are difficult to fully integrate into a lightweight backpack system.Source: §VI · p. 6
  • Backpack-mounted actuation currently uses open-loop torque control; planned end-effector feedback targets transmission-loss and hysteresis effects on torque tracking.The discussion associates output attenuation with friction, cable-path curvature, hysteresis, and pretension, and proposes optimizing pretension and routing.Source: §VI · pp. 6–7

C. Human-Worn Data Recording

  • Encoder-based hip angles remained aligned with myoMOTION, with calibrated RMSE of 5.54 ± 2.40◦ for bench-mounted walking and 3.19±1.10◦ for backpack-mounted walking.Residual differences may reflect joint-axis misalignment, soft-tissue motion, interface compliance, or different indoor and outdoor environments.Source: C. Human-Worn Data Recording · p. 6, Figure 8 · p. 7

VI. DISCUSSION

  • The lower backpack-mounted comfort score was mainly associated with the larger power supply used for outdoor recording endurance.Proposed next steps include reducing battery and power-electronics mass, improving load distribution, and refining the waist-back enclosure.Source: §VI · p. 7

VII. CONCLUSIONS

The conclusion presents a reconfigurable platform in which bench- or backpack-mounted actuators drive the same cable-free hip interface. Human-worn experiments validated configuration replacement and sensing records, while further rehabilitation and outdoor assistance studies remain planned.

VII. CONCLUSIONS

  • The platform lets bench-mounted and backpack-mounted actuation units drive the same cable-free wearable hip interface, retaining body attachments and measurement arrangement when the actuator is replaced.Quick-release end-effectors integrate torque sensing and angle measurement, with wireless IMUs and compatible sensor and controller connections.Source: §VII · p. 7
  • Human-worn experiments validated actuation-unit replacement, bench-mounted admittance-based motion tracking, backpack-mounted open-loop torque tracking, and continuous joint-motion and interaction-torque records.The work positions the platform as reusable hardware spanning laboratory prototyping, rehabilitation training, and mobile-assistance validation.Source: §VII · p. 7
  • Future work will use bench-mounted experiments for gait correction and admittance interaction, and backpack-mounted tests for outdoor active assistance and longer-term wearable evaluation.Planned outdoor tests combine gait-phase estimation, trajectory planning, and torque control.Source: §VII · p. 7
Loading 2609.25639v1…