The project centers on the creation of a cost-effective, multimodal instrumented insole designed to capture real-time gait data for controlling assistive lower-limb exoskeletons. The physical hardware consists of a flexible, 3D-printed thermoplastic polyurethane base embedded with five force-sensing resistors (FSRs) strategically placed under key anatomical landmarks of the foot. This plantar pressure data is paired with kinematic data from an inertial measurement unit (IMU) housed on a custom circuit board strapped to the user's calf. Driven by a Raspberry Pi and a lithium polymer battery, the data acquisition is synchronized into a fully wearable unit that is highly affordable.  

To interpret the raw sensor signals, the project engineered a predictive algorithmic pipeline. The system utilizes Random Forest machine learning models to identify real-time gait phases (stance versus swing) and Long Short-Term Memory (LSTM) neural networks to predict the wearer's Center of Pressure (CoP) trajectories in multiple directions. By validating this setup on human participants against gold-standard laboratory equipment, the project successfully proved that a lightweight, low-sensor-count wearable can rival the accuracy of traditional lab instrumentation. Ultimately, this framework was developed to provide the precise, real-time feedback necessary to safely actuate and synchronize robotic exoskeletons with human movement outside of clinical settings