Theses supervised by Prof. Dr. Çağatay Başdoğan
17 theses · Koç University
Robot-assisted drilling on curved surfaces with haptic guidance under adaptive admittance control
Drilling a hole on a curved surface with a desired angle is prone to failure when done manually, due to the difficulties in drill alignment and also inherent instabilities of the task, potentially causing injury and fatigue to the workers. On the other hand, it can be impractical to fully automate such a task in real manufacturing environments because the parts arriving at an assembly line can have various complex shapes where drill point locations are not easily accessible making automated path planning difficult. In this work, an adaptive admittance controller with 6 degrees of freedom is developed and deployed on a KUKA LBR iiwa 7 cobot such that the operator is able to manipulate a drill mounted on the robot with one hand comfortably and open holes on a curved surface with haptic guidance of the cobot and visual guidance provided through an AR interface. Real-time adaptation of the admittance damping provides more transparency when driving the robot in free space while ensuring stability during drilling. After the user brings the drill sufficiently close to the drill target and roughly aligns to the desired drilling angle, the haptic guidance module fine tunes the alignment first and then constraints the user movement to the drilling axis only, after which the operator simply pushes the drill into the workpiece with minimal effort. Two sets of experiments were conducted to investigate the potential benefits of the haptic guidance module quantitatively (Experiment I) and also the practical value of the proposed pHRI system for real manufacturing settings based on the subjective opinion of the participants (Experiment II). The results of Experiment I conducted with 3 naïve participants, show that haptic guidance improves task completion time by 26% while decreasing human effort by 16% and muscle activation levels by 27% compared to no haptic guidance condition. The results of Experiment II, conducted with 3 experienced industrial workers, show that the proposed system is perceived to be easy to use, safe, and helpful in carrying out the drilling task.
Design and development of a handheld haptic device for force and stretch feedback in virtual environments
This thesis presents the design and development of a handheld haptic device capable of delivering force feedback to users for grasping and squeezing virtual objects and also stretch feedback for rendering inertial effects in virtual environments. The device utilizes cable-driven mechanisms and is equipped with four force sensors and two DC motors. The design and development process involved evaluating various mechanical and electronic components, selecting the proper ones and assembling them, and conducting user studies to assess the effectiveness of the device in virtual environments.
Deep reinforcement learning to optimize task performance in human-robot co-manipulation
We propose a two-layer machine learning (ML) approach, which utilizes an artificial neural network (ANN) model as a precursor for a deep reinforcement learning (DRL) model, to optimize task performance during human-robot co-manipulation of heavy objects. In the first layer, the ANN model estimates the human intention to accelerate or decelerate the object (which precedes the actual acceleration or deceleration of the object due to its large inertia). This probabilistic estimation is then used to calculate the gain of an adaptive admittance controller, which alters the robot's contribution to the task. In the second layer, the DRL model fine-tunes this gain and optimizes the task performance by minimizing the jerk in movement and the physical effort made by human. Since online training of a DRL model for a physical human-robot interaction (pHRI) task is highly time-consuming and can potentially be dangerous for the human due to abrupt changes in controller gain, a data-driven human force model was developed by a conditional variational auto-encoder (C-VAE) for offline training of the DRL model via simulations. For this purpose, experimental data was collected from six subjects under 3 different fixed gains of the admittance controller (minimum, nominal, and maximum) to train and validate the DRL model. The adaptive admittance gain profiles generated by the ANN model alone and the proposed two-layer approach (ANN + DRL) were compared through co-manipulation simulations. The results show that the gain profile obtained by the two-layer approach leads to a decrease in human effort and jerk compared to the initial profile provided by the ANN model.
Electro-mechanical contact interactions between human finger and touchscreen under electroadhesion
Electroadhesion is a promising technology with potential applications in robotics, automation, space missions, textiles, tactile displays, and some other fields where efficient and versatile adhesion is required. However, a comprehensive understanding of the physics behind it is lacking due to the limited development of theoretical models and insufficient experimental data to validate them. In this thesis, we have developed an electro-mechanical model to estimate the magnitude of electrostatic forces between human finger and touchscreen under electroadhesion. We also measured the friction forces between the finger and touchscreen to infer the magnitude of electrostatic forces experimentally. The model is in good agreement with the experimental data and showed that the change in magnitude of the electrostatic force is mainly due to the leakage of charge from the Stratum Corneum layer of the skin to the touchscreen at frequencies lower than 250 Hz and electrical properties of the Stratum Corneum at frequencies higher than 250 Hz. In addition, we proposed a new and systematic approach based on electrical impedance measurements, where skin and touchscreen impedances are measured and subtracted from the total impedance to obtain the remaining impedance in order to estimate the electrostatic forces between the finger and the touchscreen. This approach also marks the first instance of experimental estimation of the average air gap thickness between human finger and voltage-induced capacitive touchscreen. Moreover, the effect of electrode polarization impedance on electroadhesion was investigated. Precise measurements of electrical impedances confirmed that electrode polarization impedance exists in parallel with the impedance of the air gap, particularly at low frequencies, giving rise to the commonly observed charge leakage phenomenon in electroadhesion. We also investigated tactile perception by electroadhesion for DC and AC voltage signals applied to the touchscreen using ten participants with varying finger moisture levels. Our study showed that the voltage detection threshold for an AC signal was significantly lower than that of the corresponding DC signal and we explained this discrepancy by charge leakage at lower frequencies again. We have also observed that the participants with a moist finger had significantly higher threshold levels than the rest of the participants, which is supported by our electrical impedance measurements. Finally, we aimed to investigate the effect of touchscreen's top coating layer on our tactile sensing with and without electroadhesion, and within the time frame of this thesis, we have focused on the latter only. Hence, we first performed psychophysical experiments to quantify human tactile discrimination ability of touchscreen surfaces coated with different materials, followed by multiple physical measurements. The results showed that coating material has a strong influence on our tactile perception and human finger is capable of detecting differences in surface chemistry due to, possibly, molecular interactions. In conclusion, the findings of this thesis provide new insights into the physics of finger-touchscreen interactions under electroadhesion and have implications for the design of robotic systems and haptic interfaces utilizing this technology.
Robotic learning of haptic skills from expert demonstration for contact-rich manufacturing tasks
In this study, we propose a learning from demonstration (LfD) approach that utilizes an interaction (admittance) controller and two force sensors for the robot to learn the force applied by an expert from demonstrations in contact-rich tasks such as robotic polishing. Our goal is to equip the robot with the haptic expertise of an expert by using a machine learning (ML) approach while providing the flexibility for the user to intervene in the task at any point when necessary by using an interaction controller. The utilization of two force sensors, a pivotal concept in this study, allows us to gather environmental data crucial for effectively training our system to accommodate workpieces with diverse material and surface properties and maintain the contact of polisher with their surfaces. In the demonstration phase of our approach where an expert guiding the robot to perform a polishing task, we record the force applied by the human and the interaction force via two separate force sensors for the polishing trajectory followed by the expert to extract information about the environment. An admittance controller, which takes the interaction force as the input is used to output a reference velocity to be tracked by the internal motion controller (PID) of the robot to regulate the interactions between the polisher and the surface of a workpiece. A multilayer perceptron (MLP) model was trained to learn the human force profile based on the inputs of Cartesian position and velocity of the polisher, environmental force, and friction coefficient between the polisher and the surface to the model. During the deployment phase, in which the robot executes the task autonomously, the human force estimated by our system is utilized to balance the reaction forces coming from the environment and calculate the force needs to be inputted to the admittance controller to generate a reference velocity trajectory for the robot to follow. We designed three use-case scenarios to demonstrate the benefits of the proposed system: we first compare the performance of an expert polisher with a naive user to show the importance of haptic skills in polishing. In the second use-case, we show that the proposed system can successfully learn the intrinsic changes in human force profile from the expert user for autonomous robotic polishing of workpieces of different material and surface properties. The last use-case scenario involves human intervention during the robotic polishing for the regions on the workpiece requiring more polishing. The presented use-cases highlight the capability of the proposed pHRI (physical Human Robots Interaction) system to learn from human expertise and adjust its force based on material and surface variations during automated operations, while still accommodating manual interventions as needed.
Effect of finger orientation on contact stiffness and area during sliding on a smooth surface
Earlier experimental studies showed that the apparent contact area of human fingerpad shrinks and eventually reaches a steady-state value as it slides on a smooth surface, although the root causes of this reduction have not been fully understood yet. We hypothesize that finger rotation about its axial axis and the movement direction play critical roles in the area change. To test this hypothesis, we conducted an experimental study to investigate the evolution of apparent contact area between human fingerpad and a smooth flat surface under normal loading (stationary finger) and combined loading (sliding finger) conditions for 4 different internal rotations of the index finger (away from the second finger) about its axial (longitudinal) axis and 2 different sliding directions. Our results show a reduction in the contact area for radial sliding as expected, but a surprising increase in the ulnar direction for the higher finger rotations. We argue that this asymmetric behavior in contact area evolution stems from the changes in the equivalent radius of curvature and stiffening of the finger as the rotation angle increases, which manifests itself as the asymmetric stress distribution at the leading and trailing edges of the fingerpad in our finite element simulations.
A machine learning approach to resolving conflicts in physical human-robot interaction
As artificial intelligence techniques become more sophisticated, we anticipate that robots collaborating with humans will develop their own intentions, leading to potential conflicts in interaction. This development calls for advanced conflict resolution strategies in physical human-robot interaction (pHRI), a key focus of our research. We use a Machine Learning (ML) classifier to detect conflicts during co-manipulation tasks to adapt the robot's behavior accordingly using an admittance controller. In our approach, we focus on two groups of interactions, namely "harmonious" and "conflicting", corresponding to the cases of the human and the robot working in harmony to transport an object when they aim for the same target and human and robot are in conflict when human changes the manipulation plan such as a change in the direction of movement or parking location of the object, respectively. Co-manipulation scenarios were designed to investigate the efficacy of the proposed ML approach, involving 20 participants. Task performance achieved by the ML approach was compared against three alternative approaches: a) a Rule-Based (RB) Approach, where interaction behaviors were rule-derived from statistical distributions of haptic features; b) an unyielding robot that is proactive during harmonious interactions but does not resolve conflicts otherwise, and c) a passive robot which always follows the human partner. This mode of cooperation is known as "hand guidance" in pHRI literature and is frequently used in industrial settings for so-called "teaching" a trajectory to a collaborative robot. The results show that the proposed ML approach is superior to the others in task performance. However, a detailed questionnaire administered after the experiments, which contains several metrics, covering a spectrum of dimensions, to measure the subjective opinion of the participants reveals that the most preferred mode of interaction with the robot is surprisingly passive. This preference indicates a strong inclination towards an interaction mode that gives more control to humans and offers less demanding interaction, even if it is not the most efficient in task performance. Hence, there is a clear trade-off between task performance and the preferred mode of interaction of humans with a robot, and a well-balanced approach is necessary for designing effective pHRI systems in the future.
Estimating human muscular fatigue in dynamic collaborative robotic tasks with learning-based models
Assessing human muscle fatigue is critical for optimizing performance in physical human–robot interaction (pHRI) tasks and mitigating safety risks for the human operator. This study presents a data-driven framework for estimating muscle fatigue in dynamic pHRI tasks using surface electromyography (sEMG) sensors attached to the human arm. Subject-specific machine learning (ML) regression models were developed to estimate fatigue levels during cyclic (i.e., repetitive) pHRI tasks. Specifically, Random Forest, XGBoost, and Linear Regression models were trained to estimate the fraction of cycles to fatigue (FCF) using three frequency-domain and one time-domain EMG features. Their performance was benchmarked against a convolutional neural network (CNN) that processes spectrogram representations of filtered EMG signals. Unlike most earlier data-driven approaches that primarily formulated fatigue estimation as a classification problem, our method models the continuous progression of fatigue through regression, enabling tracking of gradual physiological changes rather than discrete states, which is critical for timely intervention and adaptive control in dynamic pHRI tasks. Experiments were conducted with ten participants who interacted with a collaborative robot operated under an admittance controller, performing lateral (left-right) cyclic movements of the end effector until the onset of muscular fatigue. The results demonstrate that the root mean square error (RMSE) of FCF estimation across participants was 20.8 ± 4.3\%, 23.3 ± 3.8\%, 24.8 ± 4.5\%, and 26.9 ± 6.1\% for the CNN, Random Forest, XGBoost, and Linear Regression models, respectively. To examine cross-task generalization in this investigational study, additional experiments were performed with one participant who executed vertical (up–down) and circular repetitive movements. Models trained solely on the lateral-movement data were directly tested on these unseen tasks. The results indicate that the proposed ML/DL models are robust to variations in movement direction, arm kinematics, and muscle recruitment patterns, while the Linear Regression model performed poorly.
Piezo tahrikli dokunmatik bir ekranda görüntülenen sanal tümseklerin dokunsal algısı
This thesis aims to design a piezo-actuated touch screen for displaying vibrotactile haptic feedback. We have designed a touch screen to display haptic bumps to a user via four rectangular piezo patches attached to its surface. Using a finite element (FE) model, the dimensions of the screen and the appropriate boundary conditions are determined carefully based on the range of vibration frequencies detectable by a human finger. Then, the optimum configuration (location and orientation) for the piezo patches is determined such that the vibration amplitude of the screen is maximized for a unit voltage applied to each piezo patch. Our FE analysis shows that the best output is obtained when the piezo actuators are placed close to the boundaries and not the free edges. In order to validate the results obtained through the FE model, an experimental modal analysis of the glass plate is also performed using a laser Doppler vibrometer (LDV). Following the design and characterization of the plate, we have investigated the effect of displaying haptic bumps through the plate on human perception of roughness by conducting psychophysical experiments with 10 subjects. To generate the desired haptic bumps displayed to the subjects, the high frequency vibrations of the plate at its first resonance frequency is modulated with desired low frequency unipolar pulse waves in different shape (sinusoidal, square, sawtooth), peak amplitude, duty cycle, and spatial frequency. The subjects are asked to rate their tactile perception of 81 different haptic bumps on a Likert scale of 1-7 using the adjectives of smooth-rough and flat-bumpy. The results of the user study show that the perceived roughness increases with increasing peak amplitude, duty cycle, and spatial frequency. The subjects perceived the square wave as the roughest while the sawtooth was perceived as the smoothest among 3 waveforms. It is also observed that there is a close relation between the root mean square (RMS) values of the signals and roughness perception. Perception of roughness increases as the RMS values of the signals increases.
Koruma solüsyonunun ve zamanının karaciğerin mekanik ve histolojik özelliklerine etkisi
In liver transplantation, most of the time, the donor and recipient are in different locations and longer preservation periods are inevitable. Hence, the choice of the preservation solution and the duration of the safe preservation period are critical for the success of the transplant surgery. In this thesis, we examine the mechanical and histological properties of bovine liver stored in Lactated Ringer's, HTK and UW solutions as a function of preservation period to investigate the efficiencies and safe preservation periods of these solutions. The mechanical experiments are conducted with a shear rheometer on cylindrical tissue samples extracted from three bovine livers and the change in viscoelastic material properties of the bovine liver is characterized using the fractional derivative Kelvin–Voigt Model. Also, the histological examinations are performed on the same liver samples under a light microscope. The results show that the preservation solution and period have a significant effect on the mechanical and histological properties of the liver tissue. The storage and loss moduli, the number of the apoptotic cells, the collagen accumulation, and the sinusoidal dilatation increase, and the glycogen deposition decreases as the preservation period is increased. Based on the statistical analyses, we observe that the liver tissue is preserved well in all solutions for up to 11 h. After then, UW solution provides a better preservation up to 29 h. However, for preservation periods longer than 29 h, HTK is a more effective preservation solution if the least amount of change in mechanical properties is considered as the criteria. On the other hand, the highest correlation between the mechanical and histological properties is observed for the liver samples preserved in UW solution.
Reolojik ölçümlerde normal gerinimin kayma modülü üzerine etkisi
In experiments performed to characterize rheological properties of viscoelastic solids, some amount of compression is necessary to prevent the slippage between the sample and the parallel plates of a rheometer. While the effect of normal compression on the shear rheological properties of viscoelastic materials has been already acknowledged and reported in few studies in the literature, to our knowledge, no systematic study has been conducted to investigate this effect in detail to date. In this study, we perform two sets of experiments to investigate the effect of normal strain and strain rate on the dynamic shear moduli of bovine liver. First, we apply normal compressive strain to the cylindrical bovine samples up to 20% at loading rates of v = 0.000625, 0.00625, 0.0625, 0.315, 0.625 mm/s. Second, we perform torsional shear loading experiments in the frequency range of ω = 0.1 to 10 Hz under varying amounts of compressive pre-strain (ε = 1%, 2.5%, 5%, 7.5%, 10%, 12.5%, 15%, 17.5% and 20%) applied at the quasi-static loading rate of v = 0.000625 mm/s. The results of the experiments show that the shear moduli of bovine liver increase with compressive pre-strain. A hyper-viscoelastic constitutive model is developed and fit to the experimental data to estimate the true shear moduli of bovine liver for zero pre-compression. With respect to this reference value, the mean relative error in measurement of shear moduli of bovine liver varies between 0.2% and 243.1% for the compressive pre-strain varying from ε = 1% to 20%. The dynamic shear modulus of bovine liver for compressive pre-strain values higher than ε = 2.5% are found to be statistically different than the true shear moduli estimated for zero compressive strain (p < 0.05).
Sanal dokuların ultrasonik eyleme tekniği ile titretilen bir ekranda dokunsal görüntülenmesi
To render tactile cues on a touchscreen by friction modulation, it is important to understand how human perceive a change in friction. This thesis aims to explore the tactile perception of changes in friction (called edges), and underlying perceptual mechanism. We first estimated the perceptual thresholds to detect a virtual edge, rendered by rising friction (RF) and falling friction (FF), while finger is sliding on the touch surface. Then, we conducted intensity scaling experiments and investigated the effect of finger sliding velocity, normal force, and rise/fall time of vibration amplitude (transition time) on the perceived intensity of a virtual edge. In order to better understand the role of contact mechanics, we also looked into the correlations between the perceived intensities of subjects and several parameters involved in contact. The results of our experiments showed that the contrast and rate of change in tangential force were best correlated with the perceived intensity. The subjects perceived RF more strongly than FF, particularly at higher tangential force contrast. The results also showed that transition time and normal force have significant effect on our tactile perception. In the second part, we investigated our ability to discriminate two consecutive edges, followed by discrimination and roughness perception of multiple edges (called periodic gratings). The results showed that discrimination of two consecutive edges was significantly influenced by edge sequence: FF followed by RF was discriminated more easily than the reverse order. On the other hand, periodic gratings generated by displaying consecutive edge sequence of FF followed by RF were perceived with the same acuity as compared to vice versa. We found that a relative difference of 14% in spatial period was required to discriminate two periodic gratings independent of the edge sequence. Moreover, the roughness perception of periodic gratings decreased with increasing spatial period for the range that we have investigated (spatial period > 2 mm), despite the lack of spatial cues on texture height. We also observed that rate of change in friction coefficient was better correlated with the roughness perception than the friction coefficient itself. These results will further help to understand and design of virtual textures for touch surfaces.
Elektrotitreşim için kontak mekaniği
The understanding and control of human skin contact against technological substrates is the key aspect behind the design of several electro-mechanical devices. Among these, surface haptic displays that modulate the friction between the human finger and touch surface are emerging as new user interfaces. One such modulation can be achieved by applying an alternating voltage to the conducting layer of a capacitive touchscreen, in order to control electroadhesion between its surface and the finger-pad. However, our knowledge and understanding on the underlying contact mechanics is very limited. We conducted an experimental study to investigate the contact mechanics between the human finger and a touchscreen under electrovibration using a robotic set-up and an imaging system. The results show that the coefficient of friction increases under electrovibration as expected, but the apparent contact area is significantly smaller during full slip when compared to that of no electrovibration condition. We speculate that the main cause of the increase in friction is an increase in real contact area and the reduction in apparent area is due to stiffening of the finger skin in tangential direction. We also present the adverse effect of fingertip skin moisture on electrovibration. We speculate that high moisture decreases the electrical impedance of the interfacial gap between the finger and the touchscreen, leading to a smaller increase in frictional force under electrovibration. Finally, we use a mean field theory based on multiscale contact mechanics to investigate the effect of electroadhesion on sliding friction, and the dependency of the finger–touchscreen interaction on the applied voltage and other physical parameters.
Elektrotitreşim ile üretilen sanal ızgara yüzeylerin dokunsal pürüzlülük hissiyatı
Realistic display of tactile textures on touch screens is a big step forward for haptic technology to reach a wide range of consumers utilizing electronic devices on a daily basis. Since the texture topography cannot be rendered explicitly by electrovibration on touch screens, it is important to understand how we perceive the virtual textures displayed by friction modulation via electrovibration. We investigated the roughness perception of real gratings made of plexiglass and virtual gratings displayed by electrovibration through a touch screen for comparison. In particular, we conducted two psychophysical experiments with $10$ participants to investigate the effect of spatial period and the normal force applied by finger on roughness perception of real and virtual gratings in macro size. We also recorded the contact forces acting on the participants' finger during the experiments. The results showed that the roughness perception of real and virtual gratings are different. We argue that this difference can be explained by the amount of fingerpad penetration into the gratings. For real gratings, penetration increased tangential forces acting on the finger, whereas for virtual ones where skin penetration is absent, tangential forces decreased with spatial period. Supporting our claim, we also found that increasing normal force increases the perceived roughness of real gratings while it causes an opposite effect for the virtual gratings. These results are also consistent with the tangential force profiles recorded for both real and virtual gratings. In particular, the rate of change in tangential force (dF_t/dt) as a function of spatial period and normal force followed trends similar to those obtained for the roughness estimates of real and virtual gratings, suggesting that it is a better indicator of the perceived roughness than the tangential force magnitude.
Viskoelastik Materyallerin Dokunsal Algısı
In our daily lives, we interact with different types of deformable materials. Regarding their mechanical behavior, some of those materials lie in a range that is between purely elastic and purely viscous. This range of mechanical behavior is described as viscoelasticity. In certain types of haptic interactions such as assessment of ripeness of fruit, firmness of cheese, and consistency of organ tissue, we rely heavily on our haptic perception of viscoelastic materials. The relationship between the mechanical behavior of viscoelastic materials and our perception of them has been investigated in the field of psychorheology. However, our knowledge on how we perceive viscoelastic materials is still quite limited though some research work has already been done on purely elastic and purely viscous materials. History- and frequency-dependent behavior of viscoelastic materials results in a complex time-dependent response, which requires relatively more sophisticated models to investigate their behavior than those of purely elastic and viscous materials. In this study, we model viscoelasticity using a ``springpot'' (i.e., fractional order derivative element) and express its behavior in the frequency domain using two physical parameters: ``magnitude'' and ``phase'' of complex stiffness. In the frequency domain, we are able to devise signal detection experiments where we can investigate the perception of viscoelastic materials using the perceptual terms of ``firmness'' and ``bounciness'', corresponding to the physical parameters of ``magnitude'' and ``phase''. The results of our experiments show that the JND for bounciness increases linearly with increasing ``phase'', following Weber's law, while the JND for firmness is surprisingly independent of the level of ``phase''.
Elektrotitreşim yoluyla kademeli sürtünme değişimi
Rendering tactile effects on a touch screen via electrovibration has many potential applications. However, our knowledge on tactile perception of change in friction and the underlying contact mechanics are both very limited. In this thesis, we investigate the tactile perception and the contact mechanics for a step change in friction under electrovibration during a relative sliding between a finger and the surface of a capacitive touch screen. First, we conduct magnitude estimation experiments to investigate the role of normal force and sliding velocity on the perceived tactile intensity for a step increase and decrease in friction, called rising friction (RF) and falling friction (FF). To investigate the contact mechanics involved in RF and FF, we then measure the frictional force, the apparent contact area, and the strains acting on the fingerpad during sliding at a constant velocity under three different normal loads using a custom-made experimental setup. The results show that the participants perceived RF stronger than FF, and both the normal force and sliding velocity significantly influenced their perception. These results are supported by our mechanical measurements; the relative change in friction, the apparent contact area, and the strain in the sliding direction were all higher for RF than those for FF, especially for low normal forces. Taken together, our results suggest that different contact mechanics take place during RF and FF due to the viscoelastic behavior of fingerpad skin, and those differences influence our tactile perception of a step change in friction.
EMG sinyalleri ile eğitilmiş yapay sinir ağları kullanılarak fiziksel ̇insan-robot etkileşimi sırasında ̇insan hareketinin niyetinin tespiti
With the recent advances in cobot (collaborative robot) technology, we can now work with a robot side by side in manufacturing environments. The collaboration between human and cobot can be enhanced by detecting the intentions of human to make the production more flexible and effective in future factories. In this regard, interpreting human intention and then adjusting the controller of cobot accordingly to assist human is a core challenge in physical human-robot interaction (pHRI). In this study, we propose a classifier based on Artificial Neural Networks (ANN) that predicts intended direction of human movement by utilizing electromyography (EMG) signals acquired from human arm muscles. We employ this classifier in an admittance control architecture to constrain human arm motion along the intended direction and prevent undesired movements along other directions. The proposed classifier and the control architecture have been validated through a path following task by utilizing a KUKA LBR iiwa 7 R800 cobot. The results of our experimental study with 6 participants show that the proposed architecture provides an effective assistance to human during the execution of task and reduces undesired motion errors, while not sacrificing from the task completion time.