A Real-time Motion Tracking Wireless System for Upper Limb Exosuit Based on Inertial Measurement Units and Flex Sensors

S. S. Pastor et al. / IJE TRANSACTIONS C: Aspects Vol. 32, No. 6, (June 2019) 820-827 821 In general, a teleoperated system is the integration of a robotic device (slave) and a master device, which allows it to control the slave remotely. This class of systems are usually implemented when the task to be executed by the robot arm works autonomously presenting great complexity to itself or when the robot works in an environment that can implied a risk to the human being. In the case of robotic arms, this kind of platform are widely used to teach tasks by imitation of the human’s motion or to execute tasks that required a highly precision but claims human’s supervision as is the case of robots for surgery [8]. For these systems to be successful and to take advantage of them, it is required the development of an interface human-robot that allows an accurately control and a suitable data-feedback. An example of these interfaces thoroughly spread, with a large amount of acceptance in the research area are the haptic devices in virtual environments. Nevertheless, most of these devices are not intuitive enough to inexperienced users that do not have enough knowledge in the area of robotics [9]. On the other hand, among the most recent methods of teleoperation systems that provide a more natural integration to the user are those that allow direct mapping to the human’s joints, in this case this method has been taken in care to the development of this work 1. 1. Architecture System Figure 2 shows a functioning diagram of the teleoperation system, where it can be seen how the different elements interact based on the human being or user due to this one use the exoskeleton, which is in charge of measuring the motion of the different joints of human’s upper limb. It sends the data collected to the computer application that manage the data (angular displacement related to each joint). In order to reply the human’s motion in a 3D model of the robotic arm, this data is also sent to the microcontroller that have power over the manipulator. 1. 1. 1. Hardware Structure a) Upper-limb Exoskeleton: The implemented device, can measure the different degrees of freedom (DOF) based on the joints chosen of the human arm model stablished. This exoskeleton is lightweight, wireless and it has a good wearability being flexible. Figure 2. Functioning diagram of the teleoperation system b) Manipulator: The developed manipulator has 6 DOFs in order to be achieve any arbitrary positions and orientations due to the number of DOFs. This robot arm is called for general purpose. The distribution of the joints is similar to the human arm model stablished. c) Human Arm Model implemented: The model applied has been based on Murray’s kinematic model [9]; this one has 6 revolute DOF that are distribute as follows: three joints are located in the shoulder, one in the elbow and two in the wrist (the relation between this model and the human’s arm is shown in Figure 3) 1. 1. 2. Software Framework The upper-limb exoskeleton sends to the computer application developed, the datum of the angular displacement of each joint, in order to replicate the movements performed in real time by the human arm on a 3D model placed into the app developed and a 6 DOF robot arm. 2. MATERIALS AND METHODS The proposed exosuit is able to measure 6 DOF of human’s upper-limb which are distributed as following: shoulder (flexion-extension, abduction-adduction and internal-external rotation), elbow (flexion-extension and forearm supination-pronation) and wrist (flexionextension) according to the kinematic model of Murray and the classification of arm motions made by Gopura. The fundamental features taken into account to develop the device were portability, comfortability and adjustability, for this reason the structural material was required to be light and flexible. The design was based on the anthropometric measurements of the human body (Figure 4). 2. 1. Electronics The electronic system is composed of a portable module and a PC module, as is Figure 3. Graphical representation of the DOF related to the upper-limb Figure 4. (a) Anatomy correlation between human arm and exosuit base structure in EVA foam, (b) Exosuit with all hardware attached and fabric sleeve-mitten to fit together user arm with exosuit 822 S. S. Pastor et al. / IJE TRANSACTIONS C: Aspects Vol. 32, No. 6, (June 2019) 820-827 presented in Figure 5. Portable module uses Arduino FIO (ATmega328V running at 8MHz) as the processor to treat and calculate arm posture and joints displacement captured by bend and inertial sensors. PC module receives the information through serial port and reconstruct arm posture on a virtual application. 2. 2. Sensor location The proper location of the different sensors guarantees better measurements of the variables corresponding to angular displacements to the arm’s joints. The device utilizes 2 IMUs of 9 DOF and 2 flex sensors of 50KΩ. The first ones are used to measure the more complex joints (wrist and shoulder) and second ones sense 1 DOF (elbow flexion-extension and grasping movement). IMU 1 is positioned in the back part of the arm and above the elbow (warranting non interference with its displacement) which allows to measure the mayor portion of shifting generated by the shoulder. IMU 2 is placed on the dorsal of hand slightly below knuckles in order to sense the movements belonging to wrist, for this case, supination-pronation of forearm is associated to wrist since it has a superior impact on this joint, despite it is produced on the elbow. Since flex sensors measure the amount of bending, they are located directly on the joints, elbow and metacarpophalangeal of index finger, respectively. In Figure 6, the distribution of sensors is portrayed. 2. 3. Validation Three different experiments were carried over proposed MOCAP system. In the first one, Figure 5. System topology, where portable module corresponds to hardware directly associated to exosuit and PC module is the destiny of captured data by exosuit (sent wirelessly) to be processed and displayed into a virtual application Figure 6. Sensors' location of exosuit in the arm frame. IMUs are positioned as farther of axis of rotation of joint (which is desired to track) as it is possible, whereas, flex sensors are placed straightly over the joint surface user wearing exosuit was asked to move its arm in all directions for 40 seconds, simultaneously, samples were recorded to apply the magnetometer calibration algorithm. In the second, operator was asked to execute a pick and place task, inertial data coming from the most significant movement, in this particular case, wrist extension, was recorded to analyze fusion algorithm response (calculated angle) performance respect to individual angle estimation from gyroscope and accelerometer / magnetometer. Finally, MOCAP system functionality is evaluated by comparing the estimated angles from soft exosuit against real angles measured from a digital goniometer. In order to achieve that user had to move its arm following a lineal trajectory, which comprises a total angular displacement of 90°. 2. 4. Calibration The obtained data from experiment 1 is plotted as coordinates in Cartesian plane into axes couples ( , , ) in Figure 7. Angle measurement: Results of exosuit angle estimation performance are illustrated in Figure 8 (only two upper limb movements are plotted). Fusion filter: Results of second experiment are plotted using Matlab as is shown in Figure 9. Figure 7. Projections on the plane of magnetometer tridimensional outputs plotted as coordinates. Previous calibration process (left picture), points of plane are scattered and remarkable away from the expected geometry, however, after calibration (right piture), the points of each plane describe a perfect circle centered in the origin Figure 8. Exosuit and digital goniometer angular signal comparison in degrees. Blue line corresponds to goniometer output and orange line is the angle estimation from IMU S. S. Pastor et al. / IJE TRANSACTIONS C: Aspects Vol. 32, No. 6, (June 2019) 820-827 823 Figure 9. Complementary filter signal comparison. Red line is the gyroscope angle estimation, it describes a smooth signal that significantly diverges over the time and blue line corresponds to accelerometer/magnetometer angle estimation, whose signal has high frequency components as noise. Whereas, purple line corresponds to fusion algorithm response, which compensate the defects of previous estimations, generating an accurate angle calculation. All the values are in degrees. 3. CONTROL DESIGN In order to control any system it is necessary to identify the dynamic of the plant (system), in simple terms, characterize the plant’s response against a known input, which can be mathematical represented as a transfer function. Since exist systems that are common on industrial or laboratory environments, some transfer functions’ generic structures have been created with the purpose of synthetizing the general dynamic behavior of those mechanisms, this is the case of DC motors, whose dynamics is normally approximated to first or second order equations. Most of the time, first order expression is enough good to test control techniques. The estimated transfer function that represents the studied actuators (DC motors) is shown in Equation (1), , , (1) The above expression relates an input of voltage to an output in terms of angular velocity. Nevertheless, the current work requires as output's variable the angular position, which can be easily obtained by integrating the response of the system or seen from other perspective by multiplying the transfer function by Equation (2). , , (2) Based on plant (DC motor) mathematical expression, controllers can be designed with the aim of forcing system’s dynamics to performance according to desired parameters, settling time ( and signal overshoot percentage or , in the current case, values selected were 0,4 and 0,95 → 1%