System concept design of the knee-ankle-foot robot 

Context in source publication

Context 1

... I NTRODUCTION Due to rapid population aging, stroke becomes one of the leading causes of adult disability. Limitations in walking are an important cause of disability and morbidity. Rehabilitation through physical therapy is the main treatment for such patients to regain maximum function. Brain motor network reorganization happens with motor training after stroke [1], which is known as neuroplasticity. Task specific repetitive functional training, such as human gait training has proven to be effective in neurorehabilitation. The treadmill training with body-weight support has been used for rehabilitation of stroke patients [2] [3] and patients suffered with spinal cord injury [4]. The training resulted in improved balance, lower limb motor recovery, walking speed, and other important gait characteristics such as symmetry, stride length, and double stance time[3][5]. However, the manually assisted gait training is labor intensive and physically demanding for therapies. Therefore the availability, duration, and the frequency of training sessions are often limited. Robots can be used to provide mechanical assistance for repetitive movements. Research on robotic device for gait rehabilitation has attracted strong interest in recent years. However, most of current robotic gait training systems are integrated with fixed platforms and treadmills [6]. They are bulky, expensive, and available only to big hospitals. Yet a significant portion of neurological recovery occurs after patients have been discharged from hospitals. Such patients, still have residual gait impairment, are referred to outpatient rehabilitation centers, where compliance is often poor due to inconvenience for the visits. Therefore, there is a great need for a home-based wearable robotic system for gait rehabilitation. Such robotic gait devices should be lightweight and portable, easy to don and doff, and able to capture patient performance data. A number of researchers targeted specifically for the ankle joint to tackle the drop foot problem [7-8]. Others have tried to develop powered knee orthoses to aid the knee motion [9-11]. However, research aimed at providing active assistive torque to both the knee and ankle was very limited due to the added mechanical design complexity. The ankle- foot orthosis developed by Blaya and Herr at MIT [7] weighs 6.8lbs. The knee assistive device developed by Tibion [9] also weighs more than 4.5kg. The lightweight knee-ankle-foot orthosis (KAFO) was presented in [12]. Nut it is not portable due to the tethered operation with pneumatic actuators. Obviously, inefficient actuator and mechanism design are the key limiting factors for portable robot design. In this paper, we present an intelligent compact and modular powered knee-ankle-foot orthosis for chronic stroke patients to conduct gait rehabilitation at outpatient rehabilitation centers or at private homes. We developed a novel compact compliant actuator and linkage mechanism to achieve light-weight and modular design. The rest of this paper is organized as follows. Section II provides an overview of the system configuration of the robot. Section III introduces the novel actuator design with experimental results demonstrating its force control performance. Section IV presents the biomechanical analysis of the actuator and linkage mechanism using clinical gait data. A brief conclusion is given in section V with an outline of the future work for the robot development. II. S YSTEM C ONFIGURATION OF THE R OBOT Figure 1 shows the overall system design concept. The modular system consists of an ankle foot module and a knee module. Each module is driven with the same compact compliant force controllable linear actuator, which will be explained in detail in section III. The system can become an ankle robot by removing the knee module. It can also work as a knee robot by removing the actuator from the ankle module. It is known from human biomechanics that the range of motion of the lower limb joints is within 90 o during normal walking. Therefore, a simple rocker-slider mechanism is used to achieve a compact design. The structure of the system is made of lightweight carbon fiber composite material. The total weight for the mechanical module is estimated to be less than 4Kg. The system has a suite of sensors for gait cycle detection, muscle activation monitoring, and gait assistance control. Surface EMG sensors placed at the key muscle groups ae used to monitor the muscle activation patterns along the rehabilitation process. Joint angle sensors are used to determine the gait the kinematics. Foot pressure sensors are used to detect gait phases during the stance phase. Inertia measurement unit (IMU) that measures accelerations and angular velocities are used to detect the gait phases during swing phase. This is novel feature of the proposal system. Previous ankle robots such as the MIT ankle robot [7] or knee robots such as the Tibion [9] can only detect the gait cycle at the stance phase. The control system will determine the optimal assistive torque based on the fusion of all the sensor information and the adaptive shared control algorithm. The sensor information will also provide a quantitative measure of the rehabilitation progress based on the range of motion of the joints, the speed, and the EMG signals. To achieve a light weight and portable design, we use an embedded system as the controller and Lithium Ion batteries as the power source, which will be in a small package that can either be put around the waist or carried at the back of the user, making the overall robot system self- contained and portable. III. N OVEL A CTUATOR D ESIGN AND T ESTING Current rehabilitation robots are mainly limited by inefficient actuator design. Rehabilitation robots are intrinsically interactive robots with direct physical human robot interaction. For these robots, it is important to control of the assistive force based on the needs of individual patient [13]. Safe human-machine interaction requires compliant actuators that can achieve force control with low output impedance and back-drivability [14]. The most common compliant actuator design adopted is the Series Elastic Actuator (SEA) design [15-16]. In human friendly robotics applications, SEAs offer a range of advantages over stiff actuators, which include high force controllability and fidelity, low output impedance, back-drivability, tolerance to shock and impacts. Many different SEAs have been developed for assistive and rehabilitation robots. In [17], J. Pratt et al used the linear SEA to design an assistive knee robot. Several rotary SEAs were developed [18-20]. The performance of the SEAs largely depends on the spring constant [16]. Low stiffness spring produces high fidelity of force control, low output impedance, and reduces stiction, but also limits the force range and the force control bandwidth. High stiffness spring increases large force bandwidth, but reduces force control fidelity. In order to achieve useful output force, most current SEAs are designed with very high stiffness springs, leading to compromised force control performance, low intrinsic compliance and back-drivability, and bulky and heavy mechanical systems. Study on human biomechanics indicates that in a normal overground walking gait cycle, the average torque and power in human lower limb joints are much lower than the peak values [21]. That means the actuators for gait assistance robots only need to provide a small average torque during most of gait cycle but need to have the capability to provide higher peak torque. A series elastic actuator designed for the average assistive torque needs only a softer spring and that designed for the peak torque will need a very stiff string. We developed a novel design to overcome the above limitations of the conventional SEAs, making it most suitable for gait rehabilitation robot design. In [22], we presented the design concept and demonstrated the feasibility with simulation results. We have since developed a physical prototype and obtained promising experimental results. Fig. 2 illustrates the composition of the novel actuator. The actuator consists of a DC motor which drives a ball screw through a torsional spring assembly and a pair of spur gear. A carriage contains two linear springs is used to transmit the linear motion from the ball screw nut to the output linkage. A linear potentiometer is used to measure the displacement of the linear springs. The ...