Design and evaluation of powered lumbar exoskeleton based on human biomechanics

Mechanism architecture

From the perspective of biomechanics, this paper carries out mechanical equivalent modeling of the human spine, which is equivalent to a cylindrical joint (C) and a spherical joint (S). It is ensured that the mobility of the human lumbar spine is fully considered in the proposed design of the powered lumbar exoskeleton. Based on the establishment of equivalent human lumbar spine model and preliminary evaluation of a musculoskeletal model, the powered lumbar exoskeleton configuration is determined. The 6-SPS/SP equivalent mechanism of the powered lumbar exoskeleton includes a mobile ring, a base ring and 6 SPU branch chains with the same structure but different lengths, connecting the mobile ring and base ring.

The mechanism design, shown in Figure 1, is used to achieve the desired degrees-of-freedom while meeting the unique needs of rehabilitation. The two rings are designed by Blender software (Blender Institute, Netherlands), which can ensure that the two rings conform to the physiological curvature of the human body without scanning the human body through 3D software. The two rings provide support to the lumbar and pelvis without slipping and restricting the user’s breathing. Lateral adjustment modules and longitudinal adjustment modules are added to increase adaptability, which can meet the needs of users of different sizes. Additionally, the force buffer module is designed to absorb part of the external force, thereby avoiding secondary damage to the human spine and improving the comfort during wearing.

Figure 1:

Proposed design of powered lumbar exoskeleton.

Hardware design

The hardware design system of the powered lumbar exoskeleton is shown in Figure 2, which consists of system control board, external board, and information processing module. First, the battery pack (QSO-1206, 12 V) can provide power directly to the motor control board, and also provide power to the communication module and information processing module through voltage converter. Each motor unit has one linear motor (L16-100-150-12-P, RS Company, Shanghai, China) with a potentiometer. The potentiometer can measure the length of the actuator. The stroke is 100 mm, the reduction ratio is 150/1, the power supply voltage is 12 V, the peak power point is 200 N, the reverse driving force is 102 N, and the maximum speed is 8 mm/s. The pressure sensor (DYZ-101, DA YSENSOR, China) at the end of the actuator can measure the propulsion force. Two IMUs (BWT901CL, WIT, Shenzhen, China) detect the relative angles between the front of mobile ring and base ring.

Figure 2:

The hardware of powered lumbar exoskeleton.

In order to meet the need of different occasions for low back pain, we set the powered lumbar exoskeleton to have different training modes, namely passive mode, follow-up mode, and resistive mode. In the passive mode, the powered lumbar exoskeleton is used as a power device in order to pull the waist to move. In the follow-up mode, the powered lumbar exoskeleton acts as an assistive device to assist the wearer’s movement. In the resistive mode, the powered lumbar exoskeleton is used to provide resistance for training. A handheld device allows the user to select training motions (i.e., flexion/extension, left/right lateral bending, rotation and traction), adjust training parameters (i.e., force and angle) and switch training modes (resistive mode, follow-up mode and passive mode). It transmits commands to the system control board via Bluetooth. Additionally, the force, angle and length of the actuator can be transmitted to a PC in real time via Wi-Fi.

Display of prototype

The prototype of powered lumbar exoskeleton is shown in Figure 3. Mobile ring and base ring are placed on lumbar and pelvis respectively. The entire control module is integrated and hung on the back when wearing. The whole prototype is 2.5 kg, it is small and lightweight compared to backpack-style exoskeleton such as mobile wearable waist assist robot [7].

Figure 3:

Prototype of the powered lumbar exoskeleton. Detailed display of the prototype model (A), handheld device (B).

The simulation experiment of muscle force

Experiment environment setup

There are many human musculoskeletal models in OpenSim (4.3 Stanford), and we use the complete full-body lumbar spine model (FELS) [7], 8]. The lumbar spine mechanism (i.e., L1∼L5) and lumbar muscle groups are modeled in detail in FELS. Before the simulation experiment begins, a human-machine interaction model need to be constructed. The powered lumbar exoskeleton is simplified, and the key parameters of the simplified exoskeleton (i.e., the mass, center of mass and inertia of each part) are extracted from the XML file to add to the Bodyset. The entire human-machine interaction model is shown in Figure 4A. The bushing force between the musculoskeletal model and the simplified exoskeleton is set as the spring damping model.

Figure 4:

The simulation of muscle force. The full human-machine interaction model (A). The curves of waist movement trajectories in the flexion (B). The curves of driver forces in the flexion (C), and each colour line represented the force of an actuator.

Output sensitivity and effectiveness test in the resistive mode

In the sensitivity test, a healthy participant is recruited (age: 22 years old, height: 175 cm, weight: 70 kg), as shown in Figure 5. We conduct a preliminary resistance test with wearing the powered lumbar exoskeleton during the forward flexion, left flexion and left rotation when the resistance is set to 15 N, and feedback signals from six pressure sensors are acquired. In this way, the sensitivity and effectiveness of the resistive mode are preliminarily verified.

Figure 5:

Display of wearing the powered lumbar exoskeleton. The front display of wearing the powered lumbar exoskeleton (A), the back display of wearing the powered lumbar exoskeleton (B).

The effectiveness on muscle release

The muscle forces of the low back muscle groups (i.e., iliopsoas muscle) and abdominal muscle groups (i.e., external oblique muscle) are calculated with or without the powered lumbar exoskeleton.

As shown in Figure 6, the muscle forces of left and right muscles have the same motion patterns. Compared to when no exoskeleton is worn, the muscle force of external oblique muscle with powered lumbar exoskeleton reduces significantly in the rotation (from 36 to 14 N) and lateral flexion (from 140 to 110 N). The muscle force of iliopsoas muscle decreases by about 10 N in the lateral flexion, and has no change in the rotation and flexion. The simulation results show that wearing the powered lumbar exoskeleton can alleviate muscle strain.

Figure 6:

Changes in muscle force in various motion postures with or without powered lumbar exoskeleton. The three figures represent the changes of the muscle forces of the left and right external oblique muscle in three motion posture (rotation, lateral flexion and flexion) in the top row. The three figures represent the changes of the muscle forces of the left and right iliopsoas muscle in three motion posture (rotation, lateral flexion and flexion) in the bottom row. Black colour represents wearing the exoskeleton, gray colour represents not wearing the exoskeleton, solid lines represents the right-side muscles, and dashed lines represents the left-side muscles. EO is the abbreviation for external oblique muscle, IL is the abbreviation for iliopsoas muscle.

Output sensitivity test in the resistive mode

The resistance is set to 15 N in the forward flexion, left lateral flexion and left rotation. When the body’s resistance reaches 15 N, the powered lumbar exoskeleton is driven to move in the opposite direction, as shown in Figure 7. In the forward flexion resistance, the pressures on the motors 4 and 5 are the greatest, indicating that the waist of the human is pushing back. In the left lateral flexion resistance, the pressures on the motors 5 and 6 are the greatest, indicating that the waist of the human is resisting to the right side. In the left rotation resistance, the pressures on the motors 3, 4, and 5 are the greatest, indicating that the waist of the human is resisting to the right rotation.

Figure 7:

The pressure and length of each motor in the resistive mode. The three figures represent the length of each motor in three motion posture (forward flexion, left lateral flexion and left rotation) in the top row. The three figures represent pressure of each motor in three motion posture (forward flexion, left lateral flexion and left rotation) in the bottom row. Each colour line represents an actuator.

The effectiveness on muscle release

From the simulation results shown in Figure 6, it can be observed that wearing the powered lumbar exoskeleton causes significant changes during lateral flexion movements. The muscle force of the back muscles, such as the iliopsoas, decreases by 33 %, and the muscle force of the abdominal muscles, such as the external oblique, decreases by about 20 %. During lateral rotation, the ipsilateral flexor muscles of the lumbar region are more active than the contralateral extensor muscles, and the abdominal muscles are more active than the back muscles. After wearing the exoskeleton, the external oblique muscle shows a significant decrease of about 56 %, while the iliopsoas muscle decreases less.

Although the simulation results tended to align with the functional descriptions in related literature, such as the active participation of flexor muscles while relaxation of the extensor muscles in flexion movements [11], 12], the simplifications in the model and the limitations of the simulation environment suggest that further optimization of human motion simulations deserves more in-depth exploration and analysis.

Stokes et al. [13], 14] suggested that the contraction of the abdominal muscles generated intra-abdominal pressure, which reduced the load on the lumbar region. The reduction in the force on the iliopsoas muscles aligns with their hypothesis. However, the muscle force is significantly lower than the values reported in the study by Morini et al. [12], primarily because our study calculates changes in individual muscle forces, whereas their research focused on changes in the forces of muscle groups composed of multiple muscles.

Output sensitivity test in the resistive mode

From the sensitivity test of the pressure in Figure 7, it can be seen that in the resistive mode, when the resistance force reaches the set value of 15 N, the pressure sensors can measure the output force of each actuator in real-time, enabling a real-time response in the resistive mode. However, the current resistance setting is based on the resultant moment of the powered lumbar exoskeleton, and the contact force between the powered exoskeleton and the wearer has not yet been measured.