Predictive Dynamic Simulation of Seated Start-Up Cycling Using Olympic Cyclist and Bicycle Models

Conor Jansen, John McPhee2018 · DOI

TL;DR

This paper presents a predictive dynamic simulation model of a cyclist and bicycle, crucial for optimizing track cycling performance. By combining a detailed cyclist model with a bicycle model and employing optimal control techniques, they successfully replicated experimental data for seated start-up cycling, paving the way for more sophisticated analysis of complex cycling maneuvers.

What Problem Does This Paper Address?

Optimizing athletic performance often involves understanding the intricate biomechanics of movement. For track cycling, particularly during the critical start-up phase, a deep understanding of how a cyclist and their bicycle interact dynamically is essential. Traditional experimental methods are time-consuming and can be limited in scope. Predictive computer simulations offer a powerful alternative, allowing researchers to virtually test various techniques, equipment configurations, and body positions. However, developing accurate and comprehensive models that capture the full complexity of a cyclist and their machine, especially for dynamic events like a standing start, remains a significant challenge. This paper aims to bridge this gap by creating a predictive dynamic simulation that integrates both the cyclist and bicycle, focusing on the seated start-up phase.

Methods

The researchers developed a sophisticated model using MapleSim, a multi-domain simulation software. The core of their approach involved creating a seven-degree-of-freedom, two-legged cyclist model and integrating it with a four-body bicycle model (rear wheel, front wheel, crank, and frame/fork/handlebars).

Key methodological aspects include:

  • Combined Cyclist and Bicycle Model: Unlike many previous studies that focused on either the cyclist or the bicycle in isolation, this work integrates both. The bicycle model includes a Pacejka tire model for wheel-ground contact and a simplified drivetrain with a 4:1 gear ratio.
  • Joint Torque-Driven Cyclist Model: The cyclist's lower limbs were modeled with revolute joints at the hips, knees, and ankles. Instead of modeling individual muscles, the model uses active joint torques, scaled to represent the strength of Olympic track cyclists. These torques are further modulated by joint angle and angular velocity to mimic muscle force-velocity and force-length relationships. Passive joint torques were also included.
  • Optimal Control for Prediction: To solve the predictive simulation problem, the authors employed GPOPS-II, a direct collocation optimal control software. This method converts the differential algebraic equations (DAEs) governing the system into ordinary differential equations (ODEs) and then uses an interior-point optimizer to find the optimal control strategy. The objective function was to maximize the distance covered within a fixed time period, mirroring the experimental goal.

Datasets

The model was validated against experimental data collected from seven members (four male, three female) of the Canadian Olympic program participating in the team sprint event. Each cyclist performed maximal start-up cycling on an SRM ergometer, providing kinematic data (captured by an Optotrak motion capture system) and mechanical outputs (cadence, torque, and power). The researchers also used anthropometric data from de Leva [12] and parameters from previous studies for tire models and passive joint stiffness.

Key Insights and Ideas

  • Optimal Control for Realistic Cycling Dynamics: The study demonstrates that optimal control, when applied to a combined cyclist-bicycle model, can effectively predict realistic human movement for dynamic cycling tasks like start-up pedaling.
  • Importance of Integrated Modeling: The paper highlights that for certain cycling scenarios, particularly those involving significant body position changes (like standing starts), modeling both the cyclist and the bicycle is crucial for accurate predictions.
  • Seated Start-Up Challenges: The simulations reveal that a seated start-up is biomechanically challenging due to the high inertia at the crank and the need to overcome it without the benefit of body mass leverage. This explains why real-world standing starts involve cyclists rising from the seat.
  • Torque Oscillations: The model produced larger torque oscillations than observed in experimental ergometer data, potentially due to the simplified ergometer model and the torque scaling method.

Key Results

The simulations showed good agreement with experimental data for ergometer pedaling, producing similar joint angles, cadence, torque, and power profiles. While the track bike simulations also showed similar trends, some differences were observed, which the authors attribute to the greater resistance experienced on the track bike. The computation time for the track bike simulations was approximately 271 minutes on a standard multi-core processor.

Limitations and Future Work

The authors acknowledge several limitations:

  • Joint Torque vs. Musculoskeletal Model: The use of joint torques simplifies the model but doesn't capture individual muscle contributions or coordination. A full musculoskeletal model would be computationally prohibitive for optimal control.
  • Fixed Hip Constraint: The hip joints were fixed, which is a significant simplification. In reality, hip movement is crucial for optimizing body position and generating power, especially during standing starts.
  • Standing Starts Not Modeled: The current model is for seated start-up cycling. Standing starts, a critical technique in team sprint events, involve complex upper body movements and body position shifts that are not yet incorporated.
  • Computational Cost: Optimal control simulations are computationally intensive, limiting the complexity of models that can be practically used.

Future work will focus on:

  • Incorporating Upper Body Movement: To accurately model standing starts, the model needs to include the dynamics of the head, arms, and trunk, allowing for body position adjustments.
  • Refining the Model for Standing Starts: The ultimate goal is to use this enhanced model to study the optimal full-body kinematics for standing starts, including the initial pose.
  • Further Validation: More testing with Olympic-level cyclists and their specific bicycles is needed to ensure the model's accuracy and generalizability.

Summary updated

Abstract

Predictive dynamic simulation is a useful tool for analyzing human movement and optimizing performance. Here it is applied to Olympic-level track cycling. A seven degree-of-freedom, two-legged cyclist and bicycle model was developed using MapleSim. GPOPS-II, a direct collocation optimal control software, was used to solve the optimal control problem for the predictive simulation. The model was validated against ergometer pedaling performed by seven Olympic-level track cyclists from the Canadian team. The simulations produce joint angles and cadence/torque/power similar to experimental results. The results indicate optimal control can be used for predictive simulation with a combined cyclist and bicycle model. Future work needed to more accurately model an Olympic cyclist and a standing start is discussed.
BibTeX
@misc{jansen2018predictive,
  author = {Conor Jansen and John McPhee},
  title = {{Predictive Dynamic Simulation of Seated Start-Up Cycling Using Olympic Cyclist and Bicycle Models}},
  year = {2018},
  doi = {10.3390/proceedings2060220},
}
Figure