Continuous Gait Recognition and Control of a Powered Ankle Prosthesis with Serial Elastic Actuator

Core Concepts

This paper proposes an innovative powered ankle prosthesis with a new design of planar torsional spring for a serial elastic actuator (SEA), and utilizes a multi-layered perceptron (MLP) neural network for continuous gait prediction and recognition to enable accurate and adaptive control of the prosthesis.

Abstract

The paper presents the design and development of a powered ankle prosthesis with the following key aspects: Mechanical Design: The prosthesis utilizes a combination of a motorized driver and a belt-driven transmission system with a serial elastic actuator (SEA). A new design of planar torsional spring is proposed for the SEA, which provides the desired stiffness, endurance, and potential for modification. Gait Recognition and Control: A multi-layered perceptron (MLP) neural network is used for continuous gait recognition, predicting parameters such as locomotive speed, gait phase, ankle angle, and angular velocity. The neural network is trained on gait data from musculoskeletal models and motion capture experiments, enabling accurate and adaptive prediction across a wide speed range of 0-4 m/s. The predicted gait parameters are used to control the prosthesis motor through a PID controller, enabling real-time tracking of the biological ankle behavior. Experimental Validation: Finite element analysis is conducted to optimize the design of the planar torsional spring, achieving the desired stiffness and endurance. The neural network is evaluated using metrics such as mean square error, root mean square error, and mean absolute error, demonstrating superior performance in gait recognition. Field tests on the developed prosthesis prototype show accurate tracking of biological ankle angle, torque, and power across different walking and running speeds. The proposed design and control approach provide a comprehensive solution for a powered ankle prosthesis that can effectively assist lower-limb amputees in daily activities by mimicking the biological ankle joint.

Stats

The ankle prosthesis has a mass of 2.9 kg and a motion range of ±50°. The peak torque of the prosthesis is 220.8 Nm, and the peak rotational speed is 5.65 rad/s. The motor torque constant is 0.16 Nm/A.

Quotes

"The proposed MLP configuration has 27 layers, consisting of an input layer, 12 fully-connected hidden layers, each following by a rectified linear unit (ReLU) layer, and 2 output layers, one as the middle outputs, the other as the final outputs." "For gait phase prediction of all speeds, 70% of result has a relative error within 5%, while 84% of it has a relative error within 10%." "For angle prediction, 41% of the result has a relative error of 10%, while 63% of it has a relative error of 25%. However, 87% of ankle prediction has a relative error within 10% and 96% of it has a relative error of 25%, once the angle is bigger than 15°."

Key Insights Distilled From

MLP Based Continuous Gait Recognition of a Powered Ankle Prosthesis with Serial Elastic Actuator

by Yanze Li,Fei... at arxiv.org 04-02-2024

https://arxiv.org/pdf/2309.08323.pdf

Deeper Inquiries

How can the neural network-based gait recognition be further improved to achieve even higher accuracy and robustness, especially for the prediction of angular velocity?

To enhance the accuracy and robustness of the neural network-based gait recognition system, particularly for predicting angular velocity, several strategies can be implemented: Data Augmentation: Increasing the diversity and volume of training data by incorporating various walking speeds, terrains, and user profiles can help the neural network learn a wider range of patterns and improve its generalization capabilities. Feature Engineering: Extracting more informative features from the input signals, such as incorporating higher-order derivatives of joint angles and velocities, can provide the neural network with richer information for better predictions. Model Architecture: Experimenting with more complex neural network architectures, such as recurrent neural networks (RNNs) or long short-term memory (LSTM) networks, can capture temporal dependencies in gait patterns more effectively, especially for predicting angular velocity which is influenced by previous states. Regularization Techniques: Implementing regularization methods like dropout or batch normalization can prevent overfitting and improve the network's ability to generalize to unseen data, thereby enhancing the accuracy of angular velocity predictions. Hyperparameter Tuning: Fine-tuning the hyperparameters of the neural network, such as learning rate, batch size, and activation functions, through systematic optimization techniques like grid search or random search, can help in achieving better performance. Ensemble Learning: Combining predictions from multiple neural network models can lead to more robust and accurate results, especially for complex tasks like gait recognition where different models may capture different aspects of the data. By implementing these strategies and continuously refining the neural network architecture and training process, the gait recognition system can achieve higher accuracy and robustness, particularly in predicting angular velocity.

How can the modular design of this ankle prosthesis be leveraged to develop a full lower-limb prosthetic system that seamlessly integrates the knee and ankle joints?

The modular design of the ankle prosthesis offers a foundation for developing a comprehensive lower-limb prosthetic system that seamlessly integrates the knee and ankle joints. Here are some ways to leverage this modular design: Interchangeable Components: Designing the knee and ankle joints as interchangeable modules that can be easily attached and detached allows for customization based on individual user needs and preferences. This modularity facilitates quick adjustments and replacements, enhancing user comfort and functionality. Unified Control System: Implementing a unified control system that coordinates the knee and ankle joints' movements based on gait analysis and user intent ensures seamless interaction between the two components. This integrated approach optimizes the overall performance and user experience of the lower-limb prosthetic system. Synchronized Motion: Developing communication protocols and synchronization algorithms between the knee and ankle joints enables coordinated motion during various activities like walking, running, and climbing stairs. This synchronization ensures natural and efficient movement patterns, enhancing user mobility and stability. Adaptive Functionality: Leveraging the modularity to incorporate adaptive features that adjust joint stiffness, damping, and actuation based on real-time sensor feedback enhances the prosthetic system's adaptability to different terrains and activities. This adaptiveness improves user comfort and confidence in performing daily tasks. Scalable Design: Creating a scalable design that allows for future upgrades and expansions, such as adding additional joints or functionalities, ensures the prosthetic system's longevity and relevance as technology advances. This scalability future-proofs the system and accommodates evolving user needs. By capitalizing on the modular design principles and integrating the knee and ankle joints seamlessly, a full lower-limb prosthetic system can offer enhanced mobility, comfort, and functionality to users, ultimately improving their quality of life.

What are the potential challenges and considerations in transitioning this powered ankle prosthesis design from a research prototype to a commercially viable product?

Transitioning a powered ankle prosthesis design from a research prototype to a commercially viable product involves several challenges and considerations: Regulatory Compliance: Ensuring compliance with regulatory standards and obtaining necessary certifications, such as FDA approval, is crucial for commercialization. Adhering to safety, quality, and performance regulations adds complexity and time to the product development process. Manufacturability: Optimizing the design for mass production, cost-effectiveness, and scalability while maintaining performance and quality standards is essential. Identifying reliable suppliers, manufacturing processes, and quality control measures is critical for commercial success. User Feedback and Testing: Conducting extensive user testing and gathering feedback from amputees, clinicians, and stakeholders is vital for refining the design, functionality, and user experience. Iterative prototyping and validation with end-users help ensure the prosthesis meets their needs and expectations. Market Demand and Competition: Understanding the market demand, identifying target user segments, and analyzing competitors' offerings are essential for positioning the powered ankle prosthesis effectively. Conducting market research and developing a compelling value proposition are key for successful commercialization. Cost-Benefit Analysis: Conducting a thorough cost-benefit analysis to determine the pricing strategy, production costs, profit margins, and return on investment is crucial for sustainable commercialization. Balancing affordability with advanced features and performance is essential for market acceptance. Distribution and Support: Establishing distribution channels, partnerships with healthcare providers, and after-sales support services is necessary for reaching the target market and ensuring customer satisfaction. Providing training, maintenance, and technical support enhances the product's value proposition. Intellectual Property Protection: Securing intellectual property rights through patents, trademarks, or copyrights to protect the design, technology, and innovations from competitors is essential for safeguarding the product's uniqueness and market position. By addressing these challenges and considerations thoughtfully and strategically, the transition of the powered ankle prosthesis design from a research prototype to a commercially viable product can be successful, offering innovative solutions to improve the lives of amputees.