Scholarlink

<p>Powered lower-limb prostheses aim to enhance the mobility of people with amputations in daily life. The CYBATHLON competition offers an opportunity to evaluate these devices’ performance across 10 tasks inspired by everyday activities, such as sitting and standing, avoiding obstacles, walking on uneven terrain and slopes, and climbing stairs. This study reports on the development and deployment of the CYBERLEGs X-Leg, a powered knee–ankle prosthesis equipped with a hierarchical control architecture and task-specific finite state machines (FSMs) during the CYBATHLON 2024 leg prosthesis race. The pilot, an individual with a transfemoral amputation, was trained for approximately 46&#xa0;h, attempted eight tasks during the competition, and completed three. Overall, the pilot successfully controlled the prosthesis’s behavior to execute multiple tasks. The prosthesis delivered stable knee-swing flexion during walking (<InlineEquation ID="IEq1"><EquationSource Format="TEX">\(\approx 65^\circ \)</EquationSource></InlineEquation>), obstacle-clearance trajectories driven by thigh kinematics, and knee-extension assistance during sit-to-stand (<InlineEquation ID="IEq2"><EquationSource Format="TEX">\(\approx -120\)</EquationSource></InlineEquation> Nm). The control design and competition experience underlined the importance of initiating control co-design at earlier development stages and revealed the extensive tuning effort required for a multitask FSM-based system. Performance analysis identified reliance on manual locomotion mode selection, sensitivity in threshold-based FSM transitions, and user discomfort at higher torque levels as key areas for improvement. The competition highlighted both the feasibility and the limitations of FSM-based control when deployed across diverse functional activities. These findings motivate future work on autonomous task recognition, smoother inter-state control blending, the integration of enriched sensing modalities, and human-in-the-loop optimization for user-specific personalization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cyberlegs X-Leg at CYBATHLON 2024: insights on the control development of an active transfemoral prosthesis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月26日

查看全文
  • Stefano Nuzzo,
  • Sophia Taddei,
  • Eligia Alfio,
  • Rossana Lovecchio,
  • Menthy Denayer,
  • Stijn Hamelryckx,
  • Elias Thiery,
  • Stijn Kindt,
  • María Alejandra Díaz,
  • Louis Flynn,
  • Kevin De Pauw,
  • Tom Verstraten

摘要

Powered lower-limb prostheses aim to enhance the mobility of people with amputations in daily life. The CYBATHLON competition offers an opportunity to evaluate these devices’ performance across 10 tasks inspired by everyday activities, such as sitting and standing, avoiding obstacles, walking on uneven terrain and slopes, and climbing stairs. This study reports on the development and deployment of the CYBERLEGs X-Leg, a powered knee–ankle prosthesis equipped with a hierarchical control architecture and task-specific finite state machines (FSMs) during the CYBATHLON 2024 leg prosthesis race. The pilot, an individual with a transfemoral amputation, was trained for approximately 46 h, attempted eight tasks during the competition, and completed three. Overall, the pilot successfully controlled the prosthesis’s behavior to execute multiple tasks. The prosthesis delivered stable knee-swing flexion during walking (\(\approx 65^\circ \)), obstacle-clearance trajectories driven by thigh kinematics, and knee-extension assistance during sit-to-stand (\(\approx -120\) Nm). The control design and competition experience underlined the importance of initiating control co-design at earlier development stages and revealed the extensive tuning effort required for a multitask FSM-based system. Performance analysis identified reliance on manual locomotion mode selection, sensitivity in threshold-based FSM transitions, and user discomfort at higher torque levels as key areas for improvement. The competition highlighted both the feasibility and the limitations of FSM-based control when deployed across diverse functional activities. These findings motivate future work on autonomous task recognition, smoother inter-state control blending, the integration of enriched sensing modalities, and human-in-the-loop optimization for user-specific personalization.