Abstract
In the Motion Research Group at the University of Waterloo, we investigate the dynamics, model-based control, and design optimization of multibody biomechatronic systems. Deriving the equations for these complex systems is both tedious and error-prone, so we have automated the dynamic modelling process by combining linear graph theory from mathematics with fundamental principles from physics and biology. Our symbolic computer implementation of this approach, now part of MapleSim, generates real-time simulation code and dynamic controllers for systems ranging from biomechatronic devices to autonomous vehicles. This paper discusses the advantages of a symbolic graph-theoretic approach to systems modelling and control, highlighting real-world applications including a plug-in Prius for Toyota, a rover for the Canadian Space Agency, and a stroke rehabilitation robot.
Introduction: The Complexity of Multibody Systems
As mechatronic systems become more complex—integrating mechanical, electrical, and biological components—deriving the underlying mathematical equations for their motion becomes a significant challenge. Whether it is a robot interacting with a human's musculoskeletal system or an autonomous vehicle navigating a dynamic environment, the manual derivation of these equations is a nightmare for engineers. Traditionally, this process has been tedious and highly susceptible to human error. Furthermore, even when equations are successfully derived, they are often too computationally heavy for the real-time control required in embedded systems. The research presented here addresses the need for an automated, reliable, and computationally efficient way to model these "multibody" systems across diverse domains.
The Symbolic Graph-Theoretic Approach
The core of our methodology is the use of linear graph theory as a mathematical foundation for system modeling. Instead of manually writing down Newton-Euler equations, the system's topology is represented as a graph. This allows for a unified way to describe how different components—such as links, joints, actuators, and muscles—interact.
This method is implemented symbolically using the MapleSim™ engine. Unlike numerical solvers that crunch numbers at every step, symbolic modeling generates the actual algebraic equations first. This allows for significant simplification and optimization before any code is run. The framework incorporates fundamental principles from both physics (for mechanical components) and biology (for musculoskeletal modeling), allowing for a seamless "biomechatronic" model. Finally, the symbolic engine automatically generates optimized C-code for real-time simulation and dynamic control.
Key Insights and Real-World Applications
By using symbolic math rather than pure numerics, the resulting controllers are lean enough to run in real-time on embedded hardware, which is often a bottleneck for complex multibody systems. Linear graph theory acts as a "common language" between different engineering disciplines; the mathematical structure remains consistent whether modeling a car's suspension or a human's elbow.
The versatility of this approach has been validated through several successful real-world applications:
- Automotive: Development of a model-based controller for a Toyota plug-in Prius.
- Aerospace: Modeling and control of a planetary rover for the Canadian Space Agency (CSA).
- Robotics: Creation of a "hockey slapshot robot" to test high-speed dynamics.
- Stroke Rehabilitation: One of the most compelling applications is a rehabilitation robot that uses an integrated model of the patient’s musculoskeletal system and the robot as a single unified system. This allows the controller to determine truly optimal actions tailored to the patient's specific physical state.
Conclusion and Future Work
The framework successfully generates real-time simulation code and dynamic controllers that have been validated through both modeling and experimental results. This proves the real-world viability of symbolic graph-theoretic designs for high-performance systems.
Future research will focus on scaling these model-based controllers for the high-stakes, high-speed requirements of fully autonomous driving. Additionally, we are further developing biomechatronic assistive devices, such as smart prosthetics or orthotics, which require deep integration with human physiological models to provide effective support for users with mobility challenges.
Summary updated