Zhongkui Wang

Robust real time material classification algorithm using soft three axis tactile sensor: Evaluation of the algorithm

Materials and textures identification is a desired ability for robots. Developing such systems require tactile sensors that have enough sensitivity and spatial resolution, and the computational intelligence to meaningfully interpret sensor data. This paper introduces a texture classification algorithm utilizing support vector machine (SVM) classifier. Data taken from a novel three axis tactile sensor that utilize magnetic flux measurements for transduction was used to obtain the three dimensional tactile data. Frobenius norm calculated from the covariance matrix of the above data and the mean values of the three dimensional sensor data were used as features. Palpation velocity and small vertical load variances had minimum influence on the proposed algorithm. We have compared this algorithm with two other classification methods. They are: classify using the feature spatial period that is calculated from principal frequencies of the textures/material, and classify using neural network classifier with special properties of each materials tactile signals as features. For eight classes of material, the proposed algorithm performed faster and more accurately than the comparators when the scanning velocity and the vertical load varied.

Modeling and parameter identification of rheological object based on FE method and nonlinear optimization

There are many kinds of deformable objects in our living life. Some of them such as human tissues, human organs, and food exhibit rheological behaviors when they are subject to external force. In surgery simulation and food engineering, we need to simulate or control such behaviors. In this paper, four-element model associated with finite element (FE) method was employed to model rheological deformation. This model can reach a good approximation of rheological force response when the object experience a standard strain input. An identification approach for estimating physical parameters of rheological deformation was presented based on 2D FE simulation and nonlinear optimization. This identification method aimed at minimizing the difference of force response between the simulation and experiment by using nonlinear least square method. Finally, experiments and identification results were given and both modeling and identification method were validated by comparing the results of simulation and experiments.

A 3-D Nonhomogeneous FE Model of Human Fingertip Based on MRI Measurements

A 3-D nonhomogeneous finite-element (FE) dynamic model of a primate fingertip is developed in this paper based on magnetic resonance (MR) imaging measurements for better understanding the mechanism of human finger sensation. The geometries of a human fingertip are measured using an MR system, and a series of 2-D images is obtained. Utilizing a boundary tracking method, boundaries of the fingertip and distal phalanx are tracked from each image slice and a set of boundary nodes is generated to construct a 3-D tetrahedral mesh of the fingertip. The 3-D mesh is then utilized to formulate a nonhomogeneous FE dynamic model for simulating the fingertip behaviors. The constitutive model, which consists of elastic and viscous elements, is employed to govern the dynamic behaviors of individual tetrahedral FE. The FE model is further extended to deal with contact interaction between the fingertip and an external instrument. Differing with conventional fingertip models, the model presented in this paper is able to not only better represent internal and external geometries of a human fingertip but also take the tissue viscosity into consideration. Simulation and experiments are performed with both a human finger and a fingertip phantom under different indentation operations. We found that the Voigt model can simulate the force behaviors of a fingertip phantom but has difficulty to reproduce the force relaxation behavior of a human fingertip. We have therefore introduced a parallel five-parameter physical model to solve this problem.

Physical parameter identification of rheological object based on measurement of deformation and force

There are many kinds of deformable objects in our living life. Some of them exhibit rheological behaviors when they are subject to external force, such as human tissues, human organs, and food. If we want to simulate or control such behaviors, we have to know the physical parameters of the object in advance. In this paper, we propose an approach to identify these parameters based on 2D finite element (FE) simulation and measurement of deformation and force. At first, 2D FE model used to simulate rheological deformation was described. Then, identification method was presented according to the analysis of simulation results. Identification results for simulation were also given. Finally, this method was applied to a object made of clay. Deformation and force were measured by camera and tactile sensor respectively. The identification results show the validity and effectiveness of this method.

A biomimetic soft fingertip applicable to haptic feedback systems for texture identification

Humans recognize textures using the tactile data obtained from the human somatosensory system. Recognition of textures allows humans discriminate objects and materials. Moreover, by understanding the objects or materials texture, the human intuitively estimates roughness and the friction properties of the object or the material. This ability is necessary for object manipulative tasks. Likewise artificial haptic systems too, should have the ability to encode textures and feedback those data to haptic applications such as haptic displays. In this paper a biomimetic soft fingertip sensor that can be used in above haptic systems is introduced. The fingertip has the ability to detect force and vibration modalities. We propose three features calculated from the covariance signal of two adjacent accelerometers in the fingertip to use in texture identification. The covariance signal is transformed using Discrete Wavelet Transform (DWT) and the three features mentioned below are calculated. The mean and variance of the approximate signal, and the energies of the detailed signal are chosen as features. Then, the proposed features were validate by using those in an Artificial Neural Network (ANN) to classify seven wood samples. The results showed a 65% success rate in classifying wood samples and that the proposed features are acceptable to encode textures.

Soft Gripper Dynamics Using a Line-Segment Model With an Optimization-Based Parameter Identification Method

Soft robotics is an emerging field that focuses on the development and application of soft robots. Due to their highly deformable features, it is difficult to model and control such robots. In this paper, we proposed a simplified model to simulate a fluidic elastomer actuator (FEA). The model consists of a series of line segments connected by viscoelastic joints. Pneumatic inputs were modeled as active torques acting at each joint. The Lagrangian dynamic equations were derived. An optimization-based method was proposed to identify the unknown model parameters. Experiments were conducted using three-dimensional (3D) printed FEAs. Calibration results of a single FEA showed the repeatability of the pressure actuated bending angles, and the proposed dynamic model can precisely reproduce the deformation behavior of the FEA. Grasping experiments showed that the proposed dynamic model can predict the grasping forces, which was validated by a separate experiment of grasping force measurement. The presented methods can be extended to model other soft robots.

Magnetic and Mechanical Modeling of a Soft Three-Axis Force Sensor

This paper describes the modeling of a soft three-axis force sensor. The sensor has a cylindrical cantilever beam made of silicone rubber that compresses and bends when normal and tangential forces are applied. The displacement of the beams end is calculated by measuring the change of the magnetic field emitted by a permanent magnet embedded in the soft beam, at fixed points in space. Spring theory and bending theory are used to calculate the normal and tangential force components. The normal forces calculated by the proposed model and the measured values have an error less than 5% validating the analogy of the sensor to a soft cantilever beam under compression and bending. The proposed mathematical model is simple and faster than a finite-element model, and accurately represents the non-linear behavior of the sensors physical effects to applied loads.

Disposable soft 3 axis force sensor for biomedical applications

This paper proposes a new disposable soft 3D force sensor that can be used to calculate either force or displacement and vibrations. It uses three Hall Effect sensors orthogonally placed around a cylindrical beam made of silicon rubber. A niobium permanent magnet is inside the silicon. When a force is applied to the end of the cylinder, it is compressed and bent to the opposite side of the force displacing the magnet. This displacement causes change in the magnetic flux around the ratiomatric linear sensors (Hall Effect sensors). By analysing these changes, we calculate the force or displacement in three directions using a lookup table. This sensor can be used in minimal invasive surgery and haptic feedback applications. The cheap construction, bio-compatibility and ease of miniaturization are few advantages of this sensor. The sensor design, and its characterization are presented in this work.

3D printed soft gripper for automatic lunch box packing

In this paper, we proposed a 3D printed soft robot gripper with modular design for lunch box packing. The gripper consists of a rigid base and three soft fingers. A snap-lock mechanism was designed for easy attach-detach assembly of the gripper without using screws. All components were 3D printed and the soft finger structure is based on the principle of fluidic elastomer actuator. Three finger designs, and soft gripper grasping and lifting deformable objects were investigated through finite element (FE) analysis and experiments. Results suggested that different finger designs yielded different curvature along the finger and generated different stress distribution once pressurized. The proposed gripper could grasp and lift objects with variable shapes and softnesses.

Simulation of soft fingertip deformation under contact and rolling constraints using FEM and CSM

We have developed a new model to simulate contact and rolling motion between two soft fingers and an object by using Finite Element and Constraint Stabilization Methods. This model is more efficient than the penalty method in connecting boundaries of objects through discrete models such as finite element models and particle models. We show the validity of this method through the simulation of grasping and rolling using two soft fingers described by the 2D finite element model.