The smart safety shoe is a collaborative project between Allshoes safety footwear and TU Delft, initiated in 2020. The project aims to develop a safety shoe equipped with pressure sensors and machine learning capabilities to prevent lower back pain among logistics workers. Previous efforts involved two graduation projects and a student course, resulting in a prototype capable of detecting unhealthy lifting postures. This project focused on advancing the shoe's development, evaluating two sensor layouts using high-end pressure sensing insoles, and testing these on 16 participants performing various lifting tasks. The selected layout demonstrated superior performance and was further refined for future development. The shoe, now publicly presented by Allshoes, aims to enter the market by 2025, with ongoing improvements to its machine learning model and integration of additional sensors and actuators to enhance its functionality.
The logistics and warehousing industries face significant challenges related to lower back pain among workers, primarily caused by improper lifting techniques. Existing safety shoes lack the capability to prevent such injuries actively. The project aims to develop a smart safety shoe that detects and corrects unhealthy postures to mitigate the risk of lower back pain.
Data was collected using high-end pressure sensing insoles to capture plantar pressure distribution profiles from participants performing various lifting postures. The project aimed to gather comprehensive data to analyse the effectiveness of different sensor layouts. This involved testing with 16 participants who performed specific lifting tasks to ensure a robust data set for further analysis.
Advanced pressure sensing insoles were utilised to facilitate innovative data collection methods. These insoles, equipped with 230 pressure sensors per foot, provided detailed and accurate measurements necessary for developing the smart safety shoe. The equipment allowed for precise data gathering essential for the subsequent analysis and design phases.
The project involved selecting the appropriate sensor layouts from previous studies to determine their feasibility and effectiveness. Two layouts were recreated in the xsensor software, customised to match the project's requirements, and then tested for their performance using machine learning techniques.
Comparison of the collected data was performed using machine learning algorithms. Data from the manual handling tests were processed and evaluated to identify which sensor layout provided the most accurate posture detection. The AED layout was found to be more precise in classification compared to the IPD layout, as evidenced by the results from the confusion matrix.
The design phase involved iterative testing and participant feedback to refine the smart safety shoe. Co-creation with users ensured that the final product would meet user preferences and functional requirements. This collaborative approach was crucial in developing a user-centric design that effectively addressed the problem of improper lifting postures.
Machine learning played an important role in the design phase. Supervised machine learning algorithms were used to classify the collected data into predefined posture categories. The tree algorithm was selected for its simplicity and explainability, helping to automate posture detection and improve the shoe's functionality.
This project successfully advanced the smart safety shoe concept, focusing on evaluating and refining pressure sensor layouts to detect lifting postures accurately. The selected sensor layout, validated through experiments with 16 participants, demonstrated improved performance in posture detection. Further integration of additional sensors and actuators was explored, laying the groundwork for future enhancements. The project emphasised the importance of responsible data usage, adhering to GDPR guidelines to protect user privacy. The smart safety shoe, poised for market introduction by 2025, promises to revolutionise protective footwear by actively preventing injuries. The development journey highlighted the necessity of iterative testing, user feedback, and interdisciplinary collaboration, setting a solid foundation for future innovations in ergonomic safety solutions.
