Lower limb exoskeletons are electromechanical wearables, consisting of compact high or low torque motors with embedded electronics, to aid in lower gait locomotion of the user’s body. These types of robotic systems are considered as augmentative devices or orthotic devices that either enhance the user's physical movement or provide locomotive aid to users to help those having trouble walking or climbing stairs due to a disability. According to the CDC, there are at least 13.7 percent of adults within the United States who have a mobility disability. The lower gait impairments can be caused by traumatic events or neurological disorders, such as spinal cord injury (SCIs), cerebral palsy, multiple sclerosis, Parkinson's disease, and strokes. With the amount of time spent using one’s lower gaits throughout daily tasks, gait locomotion becomes critical and impactful to the quality of life that may affect the overall physical and mental health of the body.
However, research and experiments performed in recent years still showed challenges in controlling lower limb exoskeletons within the development of adaptive control systems that provide a combination of good postural stability in the face of disturbances or personalized changes while maintaining comfort for the wearer. Controlling an exoskeleton requires a complex control system design to maintain comfort, safety, and stability for the user. Its system can consist of many subsystems in managing and controlling necessary data to synchronize human-robot movements (HRM) and human-robot interactions (HRI). These systems will also have to manage and balance between data generation, motor actuation, and data transfer while minimizing the sampling time for efficient and quick real time processes. Therefore, the control system is required to be adaptive and intelligent.
Three important aspects of the lower limb exoskeleton control system to be explored are implementation of adaptive control strategies, personalized motion control methods, and method of applying corrective strategies to enhance postural stability.
The objective of this project is to design and implement an adaptive control system for lower limb exoskeleton to personalize the walking trajectory and enhance postural stability. The design will incorporate and further advance the corrective strategies with personalized locomotion that were proven effective to a multi-joint lower limb exoskeleton. The developed system will be able to utilize these strategies and models to obtain information on the user’s lower limb movements and control the lower limb exoskeleton accordingly during natural locomotion that may face disturbances or personalized changes. A lower limb exoskeleton developed by the current research team will be used for testing this autonomous controller and obtained results will be recorded, plotted, and evaluated. If the user remains balanced during personalized changes or disturbances, the implemented control system and corrective strategies are shown to be effective in enhancing postural stability.
Development of the control system consist of hardware and software. The hardware has already been created utilizing high torque DC motors controlled from a main computer performing computational work and an ESP32 microcontroller performing communication in between.
My work consist of implementing adaptive central pattern generator (ACPG) controller algorithm for personalized locomotion and enhancing postural stability through the use of zero moment point (ZMP) and center of mass (COM) algorithms to estimate and maintain postural stability.
As of the beginning January 2024 of this year, the algorithm of CPG was implemented in C++ code language for two hips along with real time
The next step being performed in implementing postural stability algorithm, making CPG adaptive, and optimizing the controllers that include tweaking the motor's PD (proportional and derivative) controller gains and minimizing error. Optimization will also include confirming all variables written in Camel Case convention and reducing areas of the code that may use excessive computer resources.
The graph below are compares between actual and desired trajectories for both right and left hips. Note: some initial error due to the homing feature turned off.
The code is executed as an executive file ran through the command line.
Realtime feedback was setup within the loop. Multiple feedback received to the main computer and parsed. This is done only because of hardware setup with the ESP32. On the side, raspberry pi is being explored to improve this process.
Below is am example of the output after receiving motor feedback through serial, parsing based on string character identifiers and assigning to specific variables, then displaying them out onto console.
Additional supporting task were performed including motor repair and cable making.
Designed and 3D printed with carbon fiber infused PETG material. The housing designed to be press-fit tolerance with ease of access to the pins of the PCB board. A slit able to slide and hold velcro straps without unnecessary movements due to friction.
The IMUs were used for the project to implement with the ZMP and COM tracking and calculating the ZMP location. The sensors were also planned to have one on the ankle to detect heel strike to track when the swing leg changes to stance leg.
Designed motor mounts for table clamping and floor stand for showcase events and testing.