Development and Validation of a Modular Sensor-Based System for Gait Analysis and Control in Lower-Limb Exoskeletons

TL;DR

Lower-limb exoskeletons face barriers from costly biomechanical evaluation and complex real-time control. The authors present a modular, open-source sensor system combining instrumented forearm crutches and 3D-printed FSR insoles with IMUs, processed by a fuzzy-logic gait-phase estimator, and managed by a central on-board unit with BLE data flow at up to $130\ \mathrm{Hz}$. Validation against gold-standard motion capture and force plates across three participants shows high agreement for anteroposterior CoP ($r = 0.907 \pm 0.038$, $RMSE = 17.2 \pm 2.50$ mm), crutch GRFs ($r = 0.945 \pm 0.023$, $RMSE = 15.3 \pm 4.21$ N), and heel-strike timing ($\text{MAE} = 28.1$ ms, $r = 0.998 \pm 0.001$), supporting real-world applicability. By releasing open-source hardware and software, the work lowers costs and accessibility barriers, enabling broader adoption and iterative development of safe, responsive exoskeleton control in everyday environments.

Abstract

With rapid advancements in exoskeleton hardware technologies, successful assessment and accurate control remain challenging. This study introduces a modular sensor-based system to enhance biomechanical evaluation and control in lower-limb exoskeletons, utilizing advanced sensor technologies and fuzzy logic. We aim to surpass the limitations of current biomechanical evaluation methods confined to laboratories and to address the high costs and complexity of exoskeleton control systems. The system integrates inertial measurement units, force-sensitive resistors, and load cells into instrumented crutches and 3D-printed insoles. These components function both independently and collectively to capture comprehensive biomechanical data, including the anteroposterior center of pressure and crutch ground reaction forces. This data is processed through a central unit using fuzzy logic algorithms for real-time gait phase estimation and exoskeleton control. Validation experiments with three participants, benchmarked against gold-standard motion capture and force plate technologies, demonstrate our system's capability for reliable gait phase detection and precise biomechanical measurements. By offering our designs open-source and integrating cost-effective technologies, this study advances wearable robotics and promotes broader innovation and adoption in exoskeleton research.

Figures (18)

• Figure 1: Outdoors experimental setup using crutches and insoles system. The two sub-systems are integrated seamlessly with the exoskeleton while the central unit is securely mounted on the on-board computer interface of the exoskeleton in the back.
• Figure 2: Overview of System Integration and Design. a) The system components integrated onto a lower limb exoskeleton include: (1) two flexible 3D-printed insoles, (2) insole data collection units, (3) forearm crutches with load cells, (4) crutch data collection units, and (5) a central data and control unit mounted on the exoskeleton’s back. b) Simultaneous data collection is achieved through a sensor framework. Modular insoles with removable FSR inserts measure pressure at critical points: the heel, first, and fifth metatarsals. Forces are captured by load cells at the crutch tips, encased in semi-flexible sleeves. Data from the sensors, processed via ESP32-S3 Feather boards, are transmitted to a central unit using BLE. This unit, powered by a Raspberry Pi 4, parses data into a CSV file, while a fuzzy logic algorithm computes gait phases. An Android application manages communication between the central unit and peripherals, overseeing data flow and system operations.
• Figure 3: Preliminary tests for informing system design . a) Comparison of GRF components for exoskeleton assisted walking. Multiple crutch strike force data were collected by the force plate and averaged over normalized gait cycles. The mean of each component is plotted along with the standard deviation is illustrated in the shaded region. b) Pressure distribution during exoskeleton-assisted walking. Plantar pressures for 30 gait cycles were recorded using Moticon insoles featuring 14 sensors distributed across the foot. The average mean (left) and maximum (right) pressures for each sensor are denoted.
• Figure 4: Example of heel strike through fuzzy logic rule-based approach. During exoskeleton gait, the gait phase is decided through a set of rules based on the membership grades assigned to the FSRs of both insoles using membership functions. Linguistic variables and logical expressions are used in order to decide the outcome, or gait phase.
• Figure 8: Proposed controllers example. The green schematic describe the normal processes of the basic controllers, whereas the additional blue controls outline the extra functions incorporated by the safety assurance controller, within the higher level controller of the system. If the decision made from the safety check proves true (or safe) then the basic controller processes take place. Finally, the purple schematics introduce the proportional multiplier directly applied onto the PID controller of the lower level control of the exoskeleton.