VirSense: a novel haptic device with fixed-base motors and a gravity compensation system

Ahmad Mashayekhi (Faculty of Mechanical Engineering, Sharif University of Technology, Tehran, Iran)
Ali Nahvi (Faculty of Mechanical Engineering, K.N. Toosi University of Technology, Tehran, Iran)
Mojtaba Yazdani (Faculty of Mechanical Engineering, K.N. Toosi University of Technology, Tehran, Iran)
Majid Mohammadi Moghadam (Faculty of Mechanical Engineering, Tarbiat Modares University, Tehran, Iran)
Mohammadreza Arbabtafti (Faculty of Mechanical Engineering, Shahid Rajaee Teacher Training University, Tehran, Iran)
Mohsen Norouzi (Faculty of Mechanical Engineering, Sharif University of Technology, Tehran, Iran)

Industrial Robot

ISSN: 0143-991X

Article publication date: 14 January 2014

414

Abstract

Purpose

This paper aims to present the design and implementation of VirSense, a novel six-DOF haptic interface system, with an emphasis on its gravity compensation and fixed-base motors.

Design/methodology/approach

In this paper, the design and manufacture of the VirSense robot and its comparison with the existing haptic devices are presented. The kinematic analysis of the robot, design of the components, and manufacturing of the robot are explained as well.

Findings

The proposed system is employed to generate a Virtual Sense (VirSense) with fixed-base motors and a spring compensation system for counterbalancing the torques generated by the weight of the links. The fixed bases of the motors reduce the system's effective mass and inertia, which is an important factor in haptic interface systems. A novel cabling system is used to transmit the motor torques to the end-effector. The spring-based gravity compensation system causes more reduction in the effective mass and inertia.

Originality/value

This paper provides the details of the VirSense haptic device, its gravity compensation system, and a novel cabling power transmission.

Keywords

Citation

Mashayekhi, A., Nahvi, A., Yazdani, M., Mohammadi Moghadam, M., Arbabtafti, M. and Norouzi, M. (2014), "VirSense: a novel haptic device with fixed-base motors and a gravity compensation system", Industrial Robot, Vol. 41 No. 1, pp. 37-49. https://doi.org/10.1108/IR-02-2013-328

Publisher

:

Emerald Group Publishing Limited

Copyright © 2014, Emerald Group Publishing Limited


1 Introduction

A user of a haptic device controls the position and movements of the end-effector and can feel forces resulting from virtual environments (Barbagli et al., 2007). In a tele-operation surgery scenario, haptic devices are employed as master robots. Several versions of haptic devices have been developed for a variety of applications (Burdea, 2000). A four-wire driven three-DOF planar haptic device is presented by Gallina et al. (2001); design and construction of a one-DOF haptic device is elaborated by Chandrabalan (2004), and a bone cutting burr using a haptic device is implemented by Agus (2004).

Compensation of gravitational torques and forces is an important factor in design of machines to decrease the input torques and energy. The gravity compensation system can be active or passive. In active systems, gravitational torques and forces are cancelled out by actuators (Fernandez et al., 2009); but in passive systems, gravity compensation is performed using the balancing masses or springs, without any external actuators. The applications of such systems include mechanisms (Wongratanaphisan and Cole, 2008; Shin and Streit, 1991), parallel robots (Wang and Gosselin, 1998; Wang and Gosselin, 1999; Baradat et al., 2008), serial industrial robots (Fattah and Agrawal, 2006), non-industrial robots (Wyrobek et al., 2008; Ulrich and Kumar, 1991), and especially in large-workspace serial manipulators where compensation of weight torques are essential (Agrawal and Fattah, 2004a, b; Walsh et al., 1991; Schenk, 2006; Moore, 2009). Gravity compensation is very useful in rehabilitation robots. In these robots, a portion of patient's weight is cancelled out to help her/him gain better mobility (Herder et al., 2006; TeRiele and Herder, 2001). Agrawal and Fattah (2003) suggested an arm Rehabilitation system based on springs and auxiliary links to compensate for a large portion of patient's and system's weight. Later, they extended this approach to a leg rehabilitation system (Agrawal and Fattah, 2004a, b; Banala et al., 2004; Agrawal and Agrawal, 2005). In this method, the center of mass of links and force/torque mechanisms are controlled by two series of supporting springs and links. The first group of springs is connected to the centers of masses and the second series are placed between links so that in any robot configuration, the summation of gravitational and spring's potential energy remains constant allowing the robot to always remain in equilibrium.

Passive gravity compensations in haptic devices are in two forms: spring-based and mass-based. For example, the Omega 7 haptic device has parallel links attached to springs for compensation of links' gravitational forces and torques (Force Dimension, 2012). In serial haptic devices, masses are usually used for passive gravity compensation. For instance, Phantom haptic devices decrease the torques of weights using motors of the second and the third axes of the robot as counterbalance masses (Cavusoglu et al., 2002).

A similar approach is employed in bigger scale in the LHIFAM robot (Borro et al., 2004). The LHIFAM robot is designed for aeronautic maintainability. In this robot, a parallelogram mechanism is used and the overall center of mass of the links remains on the first pitch axis resulting in cancelling out the weight of the links.

Gravity compensation using counterbalance masses increases the overall moving masses and the effective inertia of the robot. This is a drawback for haptic devices and reduces transparency of the forces in the virtual environment. In spring-based gravity compensation systems, gravitational torque is compensated without increasing the effective mass and inertia of the robot. So far, spring-based gravity compensation has been used in parallel haptic devices. This paper presents a haptic device called VirSense, which is the first serial haptic device that benefits from spring-based gravity compensation allowing 95 per cent of the torque to be compensated passively by the springs and 5 per cent actively by the motors in entire workspace (Figure 1). In this patented robot (Nahvi et al., 2012), all motors are mounted in a fixed base so that the moving mass and the effective mass of the robot are reduced.

In parallel haptic devices, motors are usually fixed to the base of the robot, thus the mass of the motor's body and housing is excluded from the effective mass and inertia of the robot. For example, in Omega 7 (Force Dimension, 2001a, b), Delta (Force Dimension, 2001a, b), and Novint Falcon (2012) haptic devices, the motors of translational motions are mounted on fixed bases and do not move with the robot's end-effector. But, in serial manipulators, fixing the body of all motors is not trivial. As a result, in the majority of serial robots, only the motor of the first axis is mounted on the base and the rest of motors are moving with their related links. For example, in the Phantom haptic device, out of the three motors for the end-effectors' translational movement, only the yaw axis has a fixed motor (Cavusoglu et al., 2002).

There are some manipulators with serial configurations with fixed motors on the base. For example, in the DLR lightweight robot, the motors of the yaw and pitch axes are mounted on the robot base. Through a novel cabling system, it has minimized changes in the length of the cables as the end-effector moves in the workspace. It also solves intrusion between the cables during the yaw movement (Zinn et al., 2004). There is also another serial robot designed for small surgeries (Seibold et al., 2004) that employs fixed motors. Quigley et al. (2011) used fixed motors and wires to drive their low cost manipulator. Commercial manipulator has used a novel cabling transmission system to reach zero backlashes with low friction and low inertia (Townsend and Guertin, 1999). Also several three DOF cable driven haptic devices are designed for planar movements (Yang and Zhang, 2009; Gosselin et al., 2000). Kim et al. (2002) created a tension based force feedback haptic device which can provide seven DOF feedback.

In this paper, a novel serial haptic device is presented. All motors, including one for the yaw axis and two for the pitch axes are fixed to the base via a unique cabling system. Consequently, the effective mass and inertia of the robot is reduced. This approach allows bigger and more powerful motors and larger forces to the user. In the next section, the conceptual design and the main features of VirSense are mentioned. Kinematics, Jacobian matrix, singular points, and the workspace of VirSense are presented in Section 3. Section 4 explains removal of the motion of motor's body and its novel cable transmission system. In Section 5, gravity compensation using linear springs, optimization of springs, and the errors are explained. We then compare the payload with a few other conventional haptic devices and the effective mass and inertia of VirSense are mentioned. Finally, manufacturing of the robot and its interfaces are presented.

2 Conceptual design

The main design criteria for this haptic interface include large forces at the stylus, low effective mass and inertia, large workspace, low friction, and ease of use. In order to perform the proposed movements in all directions, six DOFs are selected, of which three DOFs are active and apply forces, and three DOFs are passive and allow any desired orientation by the user (Figure 2).

The Phantom Premium 1.5 High Force (PP1.5HF) is in the same class as the VirSense. In Phantom haptic devices, the yaw motor is fixed and the weight of pitch motors 2 and 3 compensate for the weight of the links. Although this design compensates for the gravitational torques of the links, it adds to the effective mass and inertia of the system. In VirSense, since the housings of all three motors are fixed, their inertia would not be added to effective mass and inertia of the robot, allowing more powerful and heavier motors compared with PP1.5HF.

The novel features of VirSense robot in comparison with PP1.5HF are:

  • eliminating body motions for all three motors;

  • allowing more powerful motors;

  • producing bigger instantaneous and continuous forces;

  • eliminating limitations in the physical size and weight of motors;

  • compensating 95 per cent of gravitational torques by springs in the entire workspace of the robot; and

  • having twice workspace as PP1.5HF robot's workspace.

The quantitative comparison of features of VirSense and Phantom is summarized in Table I. It should be noted that this version of VirSense was developed for feasibility study of our ideas about the mechanism, cabling system and the gravity compensation. So, the above comparisons with Phantom are just based on mechanical characteristics of VirSense.

The following observations can be extracted from Table I:

  • Although the overall length of the links in VirSense and PP1.5HF are the same, the links of VirSense have bigger rotational freedom providing a workspace twice as much as the phantom's.

  • Since the bodies of all three motors are fixed, VirSense can use more powerful motors with more torques compared with Phantom. However, using heavier motors increases the mass of non-moving parts and leads to the larger total mass of VirSense in comparison with Phantom PP1.5HF. What is more important is that VirSense has lighter moving parts.

3 Robot kinematics

In this section, the forward kinematics of the robot is described. The singular points are obtained. Also singular points and the workspace volume are calculated.

3.1 Transformation matrix

Denavit-Hartenberg parameters based on Craig's (1989) symbol are used here to calculate transformation matrices. Coordinate frames are shown in Figure 3, and a i , d i , i and α i parameters are determined according to Table II.

Using these parameters, the transformation matrix between consecutive frames is obtained. The transformation matrix from the base to the wrist is obtained by equation (1) (note that origin of z 4 frame is located in the wrist): Equation 1 which symbols C and S are abbreviations for “cos” and “sin” functions.

3.2 Jacobian matrix

The Jacobian matrix transforms the joint velocities to the end point velocity and also to transform exerted forces on the stylus to the joint torques. The last row of transformation matrix (equation (1)) shows the position of the wrist of the robot relative to the base. We show them by P x , P y , P z as follow: Equation 2 The Jacobian matrix for the linear motion of the end point is obtained as: Equation 3

3.3 Determination of the singular points

In singular points, a robot loses one or more degrees of freedom in Cartesian space and cannot exert the necessary interaction forces with the user in arbitrary directions. To determine the singular points, first the determinant of Jacobian matrix is obtained in a simplified form, from the following equation: Equation 4 The robot singular points are obtained by equating this equation to zero:

  • If the robot is fully extended or flexed, the end point will be on the workspace boundary, and 3=0, π. The mechanical design of the robot does not allow this singularity to happen as it limits 3 in the range of 20°-160°.

  • All the points where equation L 2 cos  2+L 3 cos ( 2+ 3)=0 is valid. It is equivalent to the situation where the end point is located along the yaw axis of the first joint. In this configuration, the rotation of the first axis has no effect on the end point position and is omitted from the robot's workspace.

It is to be noted that 5=0, π are also robot singular points, and should be avoided.

3.4 Robot workspace

The robot rotates about the yaw axis ( 1) from −90° to +90°. VirSense has two main links for the pitch, i.e. links II and III in Figure 2. The effective length of each of these two links is 200 mm. To have better rigidity, there is an auxiliary link parallel to link II. The rotational range for 2 is from −140° to +30°, and for 3 is from +20° to +160°. The planar workspace shape generated from the rotation of the first and second pitch joint, in the aforementioned ranges, is shown in Figure 4. In the figure, 1 is constant while 2 and 3 vary in their own specified ranges.

With a 180° rotation of the workspace around the yaw axis, shown as a dashed line in Figure 5, the 3D workspace is realized. Workspace near yaw axis and also in x<0 section in also omitted. Different 3D views of this workspace are shown in Figure 5.

Inside this workspace, one can place a rectangular cube with dimensions of 240×360×440 mm. This means that the workspace of VirSense is 240×360×440 mm3, which is twice as large as the workspace of PP1.5HF (191×267×381 mm3). The accessible workspace volume of the VirSense is also twice as large as that of PP1.5HF; while sum of the lengths of links II and III of the PP1.5HF is almost equal to those of the VirSense.

4 Fixation of motor bases

In this section, we first describe the cabling system and fixation of all three motors to the base of the VirSense robot. Then, we calculate the effect of fixed motors on the effective inertia of the robot.

4.1 Cabling system

In most serial haptic devices, motor's body and housing move with the robot. For example, in Phantom haptic device, the body and housing of the motors move with the movement of the end-effector. These mass of housing of motors add to the effective mass and inertia of the robot, which is not desirable. VirSense has been designed with three fixed-base motors. In this robot, all the three motors are fixed to the robot base and the torque is transferred through cables from the motors to the robot links. Fixing the yaw motor is relatively simple, but making the body of the pitch motors stationary is not a trivial task.

When the robot rotates about the yaw axis, the transmission cables may get in the way of each other resulting in cable entanglement and rupture. Only can a novel design resolve this problem. Rotation about the yaw axis may also change the cable lengths. As shown in Figure 6, a special pulley structure has been designed in the bottom section of the robot, directing the cables toward the first and the second pitch motors. This structure prevents the cables from rupture, entanglement, or length variation.

The parts in Figure 6 are:

  • 1-8: eight idle pulleys for directing cables; and

  • 9 and 10: helical grooves on the shafts of the first and the second pitch motors.

All idle pulleys are grooved steel shaft with two bearings at their ends, through which the power is transmitted by the steel cable. Pulleys 3, 4 and 6, 8 are used to direct the outgoing cables of the first and the second motor, respectively. To have permanent tension in the cables, the cable from each motor acts for both extension and flexion. The grooves of these pulleys are exactly aligned with the outgoing cable direction so as to prevent cable entanglement and on-passing on the grooved part of the first and second pitch motors. Because of this design, the two ends of the cables on the motor shafts will never cross over each other, hence no abrasion or entanglement would occur.

The VirSense cabling system has been designed such that the cables connected to the pitch disks will not change length with the change of the yaw angle. No interference with other cables will occur either. Pulleys 1 and 2 direct the outgoing cables coming from pulleys 3 and 4 to the yaw axis, and pulleys 5 and 7 direct the outgoing cables from pulleys 6 and 8 to the yaw axis, respectively. Thus, the grooves of pulleys 1, 2, 5, and 7 are all located exactly on the yaw axis. As shown in part c of Figure 7, idle pulleys are in three categories. The grooves of pulleys 6 and 4 are located on the green line. The grooves of pulleys 3 and 8 are located on the red line, and the grooves of pulleys 1, 2, 5, and 7 are located on the blue line which is the yaw axis.

In Figure 7, the cables of the first pitch motor pass around pulleys 3 and 4. Then, the cables pass around pulleys 2 and 1, respectively. The grooves of the pulleys 1 and 2 are in the direction of the yaw axis. Cables between these pulleys and the first pitch disks have two ends. One end is on the pulley 1 or 2 and is always in the yaw direction; another end is on the first pitch disk and rotates about the yaw axis together with the disk. Then one end of these cables is fixed on the yaw axis and another end rotates around the yaw axis; this movement causes the length of the first pitch motor's cables to remain constant during motion around the yaw axis. A similar description is valid for the cables of the second pitch motor causing those cables to remain constant in length. As a result, one end of the pitch cables that rotate with the robot around the yaw axis will remain on the lateral surface of the cones whose apexes are at the locations of pulleys 1, 2 resulting in no elongation or shortening of the cables due to the yaw rotation. Same description in valid for the cables of second pitch motor and cause those cable remain constant in length.

4.2 Effect of fixing the motors

In this section, the effect of fixing the motors on the effective inertia of the robot is analyzed. For this purpose, we consider two potential designs for the VirSense. In the first design, the motors are assumed to be fixed on the base; thus the masses of the motors' body and housing have no influence on the effective inertia; just the inertia of the rotor has an influence on the effective inertia. In the second design, the motors are assumed to be mounted on the links and their body and housing move with the end-effector similar to the design of Phantom robots. In each design, the effective inertia because of the motors are calculated and compared with each other. The mass of motors, inertia of the motor's rotor and the transformation factor between the motor's shaft and the pitch disks are denoted by M, I rotor , and n, respectively. Numerical values for these parameters are as follows (Maxon Motor, 2013): Equation 5

Design 1

In design 1, the motors' body and housing are fixed; thus the effective inertia of the motors is just because of the rotors and is calculated from this equation: Equation 6

Design 2

In design 2, the motors' body and housing are not fixed and move with the movement of the end effector. We assume that the motors are would be assembled on the links like the Phantom's one. In this situation, besides the transmitted inertia from the rotors, the mass of the body and housing have also influence on the effective mass. The effective inertia is calculated as follows: Equation 7 which here “d” is the radius of the pitch disks (Figure 2). It is seen that the effective inertia is increased. The percentage of this increase is obtained as follows: Equation 8 It means that if the motors' bodies were not fixed and they were mounted similar to the Phantom's design, the transmitted inertia of each motor would have increased by 55 per cent.

5 Gravity compensation with springs

To compensate for the weights in the entire workspace of VirSense, an innovative idea is employed. Two linear springs are used to take 95 per cent of gravitational torques applied to the joints. The remaining 5 per cent is compensated actively by the motors. As shown in Figure 2, the spring no. 1 is used to compensate for the torque in the second joint, and the spring no. 2 for the third joint of the robot. In this section, gravitational torques, positioning of the springs, spring design, and optimization are presented.

5.1 Gravitational torques at the joints

The gravitational torques applied to the second and the third joint of the robot are calculated using potential energy of the links. Figure 7 shows the links schematics. The quadrilateral ABDC is a parallelogram. Link II of the robot (AB) is connected to the pitch disk no. 1 and the auxiliary link (CD) is connected to the disk no. 2 through an intermediate link (AC). The two pitch disks have been shown in Figure 2 distinctly, but coincide in Figure 7.

L C, AB , L C, AC , L C, CD , L C, BE are the distances from the mass centers of the links to the shown locations in Figure 7 and m AB , m AC , m CD , m BE are the masses of the parts AB, AC, CD, and BE,, respectively. m BE includes the mass of the wrist and the stylus as well 2 and 3 are measured with the robot encoders. t is the angle that the link BE makes with the horizon and is obtained from t = 3+ 2. Next, the robot's total gravitational potential energy relative to the datum passing through A at the first pitch axis is obtained. It is equal to the summation of the gravitational potential energy of every single link, i.e.: Equation 9 where U AB , U AC , U CD , U BE , and U g are the gravitational potential energy of parts AB, AC, CD, BE, and the overall gravitational potential energy, respectively. Writing the potential energy in terms of link angles, we have: Equation 10 where g, L AB , L AC are the gravitational acceleration, and the lengths of links AB and AC. Note that 2 is negative in Figure 7. Rearranging the above equations in the terms of angles, we obtain: Equation 11 The torques at the joints are obtained: Equation 12

5.2 Spring positioning for gravity torque compensation

The springs no. 1 and no. 2 shown in Figure 2 are used to compensate for the torques τ 2 and τ t , respectively. Spring positioning for the spring no. 1 is shown in Figure 8.

One end of the spring no. 1 is attached to the point O on the yaw axis of the robot. The other end of this spring is attached to the point P1 on the first pitch disk such that AP 1 is parallel with AB. So, the spring rotates about the yaw axis. The spring no. 2 is mounted on the other side of the robot to avoid overlap with the first spring as shown in Figure 2. The spring no. 2 is attached from a point on the yaw disk to the point P2 on the second pitch disk. AP 2 is parallel to AC.

5.3 Spring torque

In this section, the torque of the first spring is calculated. The results are generalized for the second spring as well. Figure 9 is magnified to show the placement of the first spring. It has a free length of L 0 and stiffness of K 1. The motor rotation is transmitted to the first pitch disk through a steel cable. This rotation causes the rotation of the link AB and consequently leads to displacement of the end point.

The torque of the first spring M S1 about the first pitch axis is as follows: Equation 13 where F S represents the spring tension force, α is the angle between the lines AP 1 and P 1 O, and r 0 is the radial distance from the junction point of the spring to the disk center. The law of sines for AP 1 O triangle is as follows: Equation 14 where L S is the spring length and a is the vertical distance from the junction point of the spring to the rotation axis. From the above equation and using β=π/2+ 2, we have: Equation 15 The spring length from the Pythagorean theorem is obtained as follows: Equation 16 The spring tension force is obtained from the following equation: Equation 17 where ΔL S is the change in the spring length. Substituting equations (15) and (17) into equation (13), M S1 is obtained: Equation 18 For a better analysis of spring torque, we define the term Φ1 from the above equation: Equation 19 Since Φ1 is a function of the lengths and its value depends on the angle 2, we define Φ1* as the mean of Φ1. The average of the applied torque by the spring is: Equation 20 Equation (12) leads to: Equation 21 Thus: Equation 22 It can see that if Φ1* is constant, then the gravity torques will be fully compensated by the springs. Determination of the torque of the second spring can be simply done by substituting 2 with t and K 1 with K 2. Therefore, it can be easily shown that the torque of the second spring about the second pitch joint is as follows: Equation 23 Finally, the desired spring stiffness for this spring is obtained: Equation 24

5.4 Design of the springs

The two springs used for compensating gravitational torques are designed based on equations (22) and (24). The robot's gravitational torques and the spring's torques are shown in Figures 10 and 11. As it is evident, the springs are able to compensate for the gravitational torques closely in the both pitch joints.

5.5 Spring optimization

Optimization of the springs results in better gravity compensation. If Φ1 and Φ2 terms are constant, the linear springs willfully compensate for the gravitational torques. Since Φ1 and Φ2 vary with the pitch angles, we need to find the parameter values so that the maximum gravity compensation is achieved. The springs may behave nonlinearly in the form of F S =KL S )N. But after some inspections the linear spring case (N=1) was obtained as the most optimal case for compensating the weights. Equation (20) indicates that the less the initial length of the springs (L 0), the smaller the changes in the Φ values. As a result, the assumption that the Φ1 and Φ2 are constant, is satisfied. In terms of limit, if the initial length of the springs are equal to zero, the Φ1 and Φ2 terms will be equal to 1.

The diagram of the weight torque compensation error versus the initial length of spring and diagram of the required spring coefficient versus the initial length of spring for first pitch joint is shown in Figures 12 and 13, respectively. Figure 12 shows that by decreasing the initial length of the spring, the gravity torque compensation error goes to zero. Figure 13 shows that the required value for spring coefficient is reduced as well. On the other hand, if the initial length of the spring is reduced, the spring must have more changes in its length. This causes more elongation in the spring and pushes the spring into the plastic deformation zone. The “a” parameter shown in Figure 9 is 178 mm. If the desired relative error is 2 per cent, according to Figure 12 the initial length should be 37 mm. Also according to Figure 13, the spring's stiffness should be 182.4 N/m. In this case, the spring percentage of elongation will be 481 per cent. It means the length of the elongated spring should be about 4.8 times of its initial length. This causes entering the spring into the plastic deformation zone. Therefore, the initial length of the spring should not be considered very small, and a balance should be established between that and the compensation error. In the VirSense robot, the free length and the stiffness of the first spring is 10 cm and 330 N/m, respectively. These values for the second spring are 10 cm and 224 N/m, respectively.

The gravitational torques are compensated by the springs by up to 95 per cent of the nominal values. The remaining 5 per cent are the difference between the gravitational torque and the spring's torques. Both these terms are functions of the angles. Practically, the angles are read at any time. These differences are calculated in real time. The difference is added to the motor's torques. Thus, the entire gravitational torques of the robot will be fully compensated. It should be noted that if the compensation criterion was changed from gravity torque to the gravitational potential energy, the VirSense robot would be able to compensate it by up to 99 per cent. Then the compensation error percentage would be less than 1 per cent.

6 VirSense specifications

6.1 End-effector forces

As the cabling system allows all three motors to have fixed bases, by considering the strength and fatigue of cables, we are allowed to choose more powerful and heavier motors. There is a transformation factor to increase the applied torque to the robot links. Thus, the effective rotor inertia of the motors as seen by the end-effector is as follows: Equation 25 where n is the transformation factor of the transmission system. A big n has an advantage and a disadvantage. The advantage is that it increases the applied force to the user's hand. But, it increases the effective inertia of the robot, which is not desirable. So, there should be a tradeoff between the two opposing effects. We used a transformation factor of 14 in our design as the optimum value. The VirSense is capable of applying a maximum force of 40 N instantaneously and a continuous force of 12.5 N to the user. In comparison, The PP1.5HF haptic robot applies a maximum force of 8.5 N and a continuous force of 1.4 N.

6.2 Effective mass of the robot

The effective mass of the moving parts is 665 g, which roughly equals the 650-g mass of each motor. Had it not have fixed bases for the motors, the effective mass would have exceeded 2,600 g. The fixed bases not only decrease the mass of the moving parts, but also they decrease the overall length of the links as there is no need to place the motors and their accessories on the links. Besides, the springs used for gravity compensation have removed the need for the balancing masses. These two features of the VirSense have decreased the effective mass and the moment of inertia of the robot significantly.

7 Implementation of the VirSense

To have more structural stiffness, a double-parallel link is chosen for the pitch movements as seen in the ABDC parallelogram mechanism of Figure 7. In order to decrease the effective weight and inertia, some links of the robot were designed with hollow rods and all disks are light weighed (Figure 14).

The VirSense utilizes three DC-Brushless MAXON EC-max 40-120 W with a mass of 650 g, a radius of 40 mm and a length of 88 mm. They are more powerful, heavier, and bigger than the motors used in the PP1.5HF haptic robot.

Steel twisted cables are used to transmit power from the motors to the disks. One end of each cable is wrapped around a pulley attached to the shaft of each motor. The other end is wrapped around a pitch disk. All disks and pulleys are grooved to prevent cables from knotting, sliding, and fragmenting. Instead of using friction, the cables are fixed directly to the shaft of the motors and to the pitch disks. This technique increases the effective transmitted torque. As shown in Figure 15, a novel pulley has been designed which is connected directly to the motor's shaft. This connection is specially designed by a groove and two set screws to prevent any possibility of cracks on shaft of motor while working. The pitch cables are fixed to small holes at each end of the pulley and each occupy half of the length of the pulley. The on centricity of the motor and the pulley is guaranteed by a housing as shown in Figure 15. To access the set screws, there is a hole in the housing.

The serial data transfer between the robot and the PC is carried out through an intermediate circuit and the USB port. This board is equipped with a PIC micro controller with USB 2.0 port and 12 Mbps of transfer rate. The commands from the drivers are sent through a 12-bit A/D to the motors with the transfer rate of 142 ksample/s. Data of the six encoders are collected using an HCTCIC having a 16-bit resolution.

8 Conclusion

In this paper, the design and implementation of a novel haptic device was presented. The main features of this robot are comparable with those of the PP1.5 HF haptic robot. Because of mounting all three motors on fixed bases, it became possible to use heavier and more powerful motors resulting in bigger end-point forces. A creative spring system was introduced, which compensated 95 per cent of the gravitational torque.

This version of VirSense was just a feasibility study of our novel ideas about the mechanism, cabling system and gravity compensation. In the next generation of this robot, the design will be optimized and we are planning to run more experimental works on the next version. It is possible to manufacture the links from composite materials to decrease the effective mass and increase stiffness. Installing some adjustment pulleys to increase the tension of the transmission cables can also improve the stiffness of the robot. We have designed an adjustment mechanism and will add it to the robot in the near future.


               Figure 1
             
               The VirSense haptic device with a novel design having fixed-base motors and passive gravity compensation

Figure 1

The VirSense haptic device with a novel design having fixed-base motors and passive gravity compensation


               Figure 2
             
               Mechanical design and components of VirSense robot

Figure 2

Mechanical design and components of VirSense robot


               Figure 3
             
               The location of frames on the robot

Figure 3

The location of frames on the robot


               Figure 4
             
               Workspace generated due to the first and the second pitch joint rotation

Figure 4

Workspace generated due to the first and the second pitch joint rotation


               Figure 5
             
               3D workspace of the VirSense haptic device

Figure 5

3D workspace of the VirSense haptic device


               Figure 6
             
               2D schematic view for (a) cable path for first pitch joint; (b) cable path for second pitch joint; (c) position of idle pulleys; (d) 3D view of idle pulleys

Figure 6

2D schematic view for (a) cable path for first pitch joint; (b) cable path for second pitch joint; (c) position of idle pulleys; (d) 3D view of idle pulleys


               Figure 7
             
               The VirSense schematics and the location of the mass centers

Figure 7

The VirSense schematics and the location of the mass centers


               Figure 8
             
               Schematic view of the spring positioning for the first spring; there is also another spring (the spring no. 2) between O and Pnot shown in this picture

Figure 8

Schematic view of the spring positioning for the first spring; there is also another spring (the spring no. 2) between O and Pnot shown in this picture


               Figure 9
             
               Schematic view of placement of the first spring

Figure 9

Schematic view of placement of the first spring


               Figure 10
             
               Diagram of gravitational torque at the first pitch joint of the robot (solid line) and the torque of the first spring (dashed line)

Figure 10

Diagram of gravitational torque at the first pitch joint of the robot (solid line) and the torque of the first spring (dashed line)


               Figure 11
             
               Diagram of gravitational torque at the second pitch joint of the robot (solid line) and the torque of the second spring (dashed line)

Figure 11

Diagram of gravitational torque at the second pitch joint of the robot (solid line) and the torque of the second spring (dashed line)


               Figure 12
             
               The weight torque compensation error versus the initial length for the first pitch joint

Figure 12

The weight torque compensation error versus the initial length for the first pitch joint


               Figure 13
             
               The weight torque compensation error versus the initial length for the first pitch joint

Figure 13

The weight torque compensation error versus the initial length for the first pitch joint


               Figure 14
             
               Light-weighted disks

Figure 14

Light-weighted disks


               Figure 15
             
               Pulleys connected to the motor's shaft

Figure 15

Pulleys connected to the motor's shaft


               Table I
             
               Comparison of VirSense robot and other well-known series haptic devices

Table I

Comparison of VirSense robot and other well-known series haptic devices


               Table II
             
               D-H parameters of VirSense

Table II

D-H parameters of VirSense

About the authors

Ahmad Mashayekhi received a BS degree in mechanical engineering in 2008 from K.N. Toosi University of Technology, Iran. He received hid MS degree in mechatronics engineering in 2011 from Sharif University of Technology. He also received bronze medal in Physics Olympiad in 2002. His current research interests include robotics and haptics and he is a member of Center of Excellence in Design, Robotics and Automation (CEDRA), Department of Mechanical Engineering, Sharif University of Technology, Tehran, Iran. Ahmad Mashayekhi is the corresponding author and can be contacted at: [email protected]

Ali Nahvi received his PhD in mechanical engineering from the University of Utah. He is currently a faculty member of mechatronics and applied design groups at K.N. Toosi University of Technology and the Director of Virtual Reality Laboratory. His current research interests include driving simulators, haptics, and intelligent showcases.

Mojtaba Yazdani received MBA in entrepreneurship from University of Tehran in 2013 and BS in mechanical engineering from K.N. Toosi University of Technology in 2010. He is a Research Associate, Project Manager and Head of Design Division in Human-Robot Interaction Lab (TaarLab) in University of Tehran. His current research interests include design and control of haptic devices, driving and flight simulators and parallel robots. Presently, he is working on the development of a commercial haptic device for dentistry applications.

Majid Mohammadi Moghadam received the PhD degree in mechanical engineering in 1996 from University of Toronto, Canada. He is Professor of mechanical engineering at Tarbiat Modares University, Tehran, Iran. His current research, which focuses on applied robotics and robust H∞ control, is concerned with haptic robotics, rehabilitation robotics, inspection robotics and rough terrain mobile robot design. He is a member of the Administrative Committee of Robotics Society of Iran.

Mohammadreza Arbabtafti received a BS degree in mechanical engineering in 2002 from Isfahan University of Technology, Iran. He received his MS and PhD degrees in mechanical engineering in 2004 and 2010 from Tarbiat Modares University, Iran. He is a recipient of Khwarizmi Young Award 2008. He is currently on the faculty of Shahid Rajaee Teacher Training University. His research interest is in the area of haptics and robotics.

Mohsen Norouzi received a BS degree in electrical engineering in 2007 from University of Zanjan, Iran and MS degree in mechatronic engineering from Sharif University of Technology in 2011, Iran. He is currently in Sharif Advanced Technologies Incubatore as a research member. His major interest in research is design and control of under-actuated robots.

References

Agrawal, A. and Agrawal, S.K. (2005), “Design of gravity balancing leg orthosis using non-zero free length springs”, Mechanism and Machine Theory, Vol. 40 No. 6, pp. 693-709.

Agrawal, S.K. and Fattah, A. (2003), “Design of an orthotic device for full or partial gravity-balancing of a human upper arm during motion”, IEEE International Conference: Intelligent Robots and Systems, Las-Vegas, NV, 27-31 October, IEEE Press, Piscataway, NJ, pp. 2841-2846.

Agrawal, S.K. and Fattah, A. (2004a), “Gravity-balancing of spatial robotic manipulators”, Mechanism and Machine Theory, Vol. 39 No. 2, pp. 1331-1344.

Agrawal, S.K. and Fattah, A. (2004b), “Theory and design of an orthotic device for full or partial gravity-balancing of a human leg during motion”, IEEE Transactions on Neural Systems and Rehabilitation Engineering, Vol. 12 No. 2, pp. 157-165.

Agus, M. (2004), “Haptic and visual simulation of bone dissection”, PhD dissertation, University of Cagliari, Cagliari.

Banala, S.K. , Agrawal, S.K. , Fattah, A. , Rudolph, K. and Scholz, J.P. (2004), “Gravity balancing leg orthosis for robotic rehabilitation”, IEEE International Conference: Robotics and Automation, New-Orleans, LA, USA, 26 April-1 May, IEEE Press, Piscataway, NJ, pp. 2474-2479.

Baradat, C. , Arakelian, V. , Briot, S. and Guegan, S. (2008), “Design and prototyping of a new balancing mechanism for spatial parallel manipulators”, ASME Journal of Mechanical Design, Vol. 130 No. 7, p. - , 13 pages.

Barbagli, F. , Salisbury, K. , Spence, C. , Ho, C. and Tan, H.Z. (2007), “Haptic discrimination of force direction and the influence of visual information”, ACM Transactions on Applied Perception, Vol. 3 No. 2, pp. 125-135.

Borro, D. , Savall, J. , Amundarain, A. , Gil, J.J. , Garcia-Alonso, A. and Matey, L. (2004), “A large haptic device for aircraft engine maintainability”, IEEE Computer Graphics and Applications, Vol. 24 No. 6, pp. 70-74.

Burdea, G.C. (2000), “Haptic feedback for virtual reality”, International Journal of Design and Innovation Research, Vol. 2 No. 1, pp. 17-29.

Cavusoglu, M.C. , Feygin, D. and Tendick, F. (2002), “A critical study of the mechanical and electrical properties of the PHANToM haptic interface and improvements for high performance control”, Presence: Teleoperators and Virtual Environments Archive, Vol. 11 No. 6, pp. 555-568.

Chandrabalan, C.J. (2004), “Design and construction of 1 D.O.F haptic interface”, MSc thesis, Florida State University, Tallahassee, FL.

Craig, J.J. (1989), Introduction to Robotics: Mechanics and Control, 2nd ed., Addison-Wesley, Boston, MA.

Fattah, A. and Agrawal, S.K. (2006), “Gravity-balancing of classes of industrial robots”, IEEE International Conference: Robotics and Automation, Orlando, FL, USA, 15-19 May, IEEE Press, Piscataway, NJ, pp. 2872-2877.

Fernandez, F.A.M. , Parra-Vega, V. and Olhuin-Diaz, E. (2009), “Active mechanical compensation to obtain gravity-free robots: modeling, control, design and preliminary experimental results”, IEEE, International Conference: Robotics and Automations, Kobe, Japan, 12-17 May, IEEE Press, Piscataway, NJ, pp. 1061-1066.

Force Dimension (2001a), available at: www.forcedimension.com/omega7-overview (accessed 11 January 2013).

Force Dimension (2001b), available at: www.forcedimension.com/products#delta.x (accessed 11 January 2013).

Gallina, P. , Rosati, G. and Rossi, A. (2001), “3-DOF wire driven planar haptic interface”, Journal of Intelligent and Robotic Systems, Vol. 32 No. 1, pp. 23-36.

Herder, J.L. , Vrijlandt, N. , Antonides, T. , Cloosterman, M. and Mastenbroek, P.L. (2006), “Principle and design of a mobile arm support for people with muscular weakness”, Journal of Rehabilitation Research and Development, Vol. 43 No. 5, pp. 591-604.

Kim, S. , Hasegawa, S. , Koike, Y. and Sato, M. (2002), “Tension based 7-DOF force feedback device: SPIDAR-G”, Proceedings of the IEEE Virtual Reality 2002 (VR.02).

Maxon Motor (2013), available at: www.maxonmotor.com (accessed 12 June 2013).

Moore, B. (2009), “Dynamic balancing of linkages by algebraic methods”, doctoral thesis, Johannes-Kepler University Linz, Linz.

Nahvi, A. , Mashayekhi, A. , Moghaddam, M.M. , Tafti, M.A. , Yazdani, A.R. , Lagha, M. , Shad, S.H.M. and Khadem, M. (2012), “Design and manufacture of a novel haptic robot with fixed base motors and automatic gravity compensation”, Patent No. 1391.04.17-75817, State Organization for Registration of Deeds and Properties, Tehran, Iran.

Novint Falcon (2012), available at: www.novint.com/index.php/novintfalcon (accessed 11 January 2013).

Quigley, M. , Asbeck, A. and Ng, A.Y. (2011), “A low-cost compliant 7-DOF robotic manipulator”, Proceedings of International Conference on Robotics and Automation (ICRA).

Schenk, M. (2006), “Theory and design of statically balanced tensegrity mechanisms”, MSc thesis, Delft University of Technology, Delft.

Seibold, U. , Kuebler, B. , Weiss, H. , Ortmaier, T. and Hirzinger, G. (2004), “Sensorized and actuated instruments for minimally invasive robotic surgery”, Proceedings of Eurohaptics, Munchen, Germany, 5-7 June.

Sensable (2010), available at: www.sensable.com/haptic-phantom-premium.htm (accessed 11 January 2013).

Shin, E. and Streit, D.A. (1991), “Spring equilibrator theory for static balancing of planar pantograph linkages”, Mechanism and Machine Theory, Vol. 26 No. 7, pp. 645-657.

TeRiele, F.L.S. and Herder, J.L. (2001), “Perfect static balancing with normal springs”, paper presented at Design Engineering Technical Conferences and Computers and Information in Engineering Conference, ASME, 9-12 September.

Townsend, W.T. and Guertin, J.A. (1999), “Tele-operator slave – WAM design methodology”, Industrial Robot, Vol. 26 No. 3, pp. 167-177.

Ulrich, N. and Kumar, V. (1991), “Passive mechanical gravity compensator for robot manipulators”, International Conferences: Robotics and Automation, Sacramento, CA, 9-11 April, IEEE Press, Piscataway, NJ, pp. 1536-1541.

Walsh, G.J. , Streit, D.A. and Gilmore, B.J. (1991), “Spatial spring equilibrator theory”, Mechanism and Machine Theory, Vol. 26 No. 2, pp. 55-170.

Wang, J. and Gosselin, C.M. (1998), “Static balancing of spatial three-degree-of-freedom parallel mechanisms”, Mechanism and Machine Theory, Vol. 34 No. 3, pp. 437-452.

Wang, J. and Gosselin, C.M. (1999), “Static balancing of spatial four-degree-of-freedom parallel mechanisms”, Mechanism and Machine Theory, Vol. 35 No. 4, pp. 563-592.

Wongratanaphisan, T. and Cole, M.O.T. (2008), “Analysis of a gravity compensated four-bar linkage mechanism with linear spring suspension”, ASME Journal of Mechanical Design, Vol. 130 No. 1, pp. 011006-011008.

Wyrobek, K.A. , Berger, E.H. , Van der Loos, H.F.M. and Salisbury, J.K. (2008), “Towards a personal robotics development platform: rationale and design of an intrinsically safe personal robot”, IEEE International Conference: Robotics and Automation, Pasadena, CA, 19-23 May, IEEE Press, Piscataway, NJ, pp. 2165-2170.

Yang, Y. and Zhang, Y. (2009), “A new cable-driven haptic device for integrating kinesthetic and cutaneous display”, Proceeding of: Reconfigurable Mechanisms and Robots, IEEE, ReMAR 2009, ASME/IFToMM International Conference on.

Zinn, M. , Khtib, O. , Roth, B. and Salisbury, J.K. (2004), “A new actuation concept for human-friendly robot design”, The International Journal of Robotics Research, Vol. 23 No. 4, pp. 379-398.

Acknowledgements

The authors would like to thank Iranian Ministry of Industries and Delshid Company for their financial and administrative support. Several people helped the authors in developing this robot, especially Mr S.H. Mohtasham Shad for computer interfacing, Mr M. Lagha for mechanical design, Ms S. Chaibakhsh for detailed design, and Mr H. Naeem Abyaneh and his colleagues for tedious manufacturing tasks at the machining workshop of K.N. Toosi University of Technology.

Related articles