Modeling and Control of Upright Lifting Wheelchair
S. C. Gharooni, B. Awada and M. O. Tokhi Department of Automatic Control and Systems Engineering, The University of Sheffield, UK; [Link]@[Link]
Abstract
This study is concerned with lifting and stabilization of a wheelchair, in which torque is applied to the back motors of the wheelchair, to force enough power to lift it upwards and then stabilize itself in the upright position. To achieve this; a wheelchair model is built in the Visual Nastran software environment. The development of fuzzy control strategies for control and stabilization of the wheelchair in the upright position is investigated and verified within simulation studies. The potential benefit of fuzzy control in this context is demonstrated. Keywords: Fuzzy control, inverted pendulum, upright wheelchair.
1 Introduction
Electronic wheelchair users would often like to be lifted upwards to talk to people directly or reach high places to pick up or put things down. This operation can be performed by lifting the wheelchair upwards and then stabilizing it as an inverted pendulum. Since the manual wheelchair came into production and became more widely used, engineers from different disciplines attempted to improve the manual wheelchair in terms of reliability, practicality size, weight, performance and automation. They searched for ways in designing more comfortable wheelchairs that will not exhaust the user, and therefore automated wheelchairs came into production [1]. With simple command-action motor-driven wheelchairs were soon developed into more sophisticated control systems that can climb over pavements, go downstairs/upstairs, run over rough material such as wood and sand, lift themselves upwards, and run steadily on two wheels
970 S. C. Gharooni, B. Awada and M. O. Tokhi
while going forwards, backwards or even sideways. The iBot wheelchair available from Johnson & Johnson [2] is capable of performing all the above mentioned tasks. This study is concerned with the design of a controller that will be able to lift and stabilize a wheelchair while a weight is placed on the chair. This task requires a high control precision with reliable and robust operation, offering the following advantages [3 ]: Allows the user to speak to a walking person while looking at them directly. Gives the user the ability to reach high places to pick up or put things down. Gives the user a better feeling when moving around in the city or at home, as they are at the same height as the others.
2 Methods
A three-dimensional wheelchair model resembling the real one was developed for this study. The model and associated simulations were implemented in Visual Nastran (VN) [4 ]. A fuzzy logic (FL) [5] controller was developed in Matlab/Simulink [6], connected to Visual Nastran for simulating the wheelchair motion. The performance of the system was assessed with random disturbances. 2.1 Modeling the wheelchair The wheelchair model built in VN has most of the physical properties of a wheelchair in terms of dimensions and weight, although it represents the wheelchair in its simplest form. While building the model some design compromises have to be made, as for this stage of study not every aspect of the wheelchair would be modeled in VN. Below is a point wise comment on the model differences between the V.N. wheelchair and the real one: 1. The weight and size of motors were not accounted for in the model. This is due to the fact that the amount of torque required to lift the wheelchair is still not known and therefore choosing a motor weight, size or power would not be appropriate at this time. 2. The weight of wheels was modelled in VN as 5 kg although in the real one the normal wheel would not weigh more than 2 kg. This is due to the following two reasons:
Modeling and control of upright lifting wheelchair
971
To stabilize the wheelchair because light wheels would make the wheelchair unstable, and difficult to keep on the ground. To account for the un-modelled weight of the motors, as eventually the weight of motors will have to be added. The motors will be connected to the axis and the wheel, which means that any extra motor weight will directly influence the wheels stability and grip to the ground. To account for the weight of the human body on the wheelchair, a humanoid model in full weight scale was put on the chair at the centre of the seat, so that its weight is acting directly downwards. Fig. 1 shows the wheelchair model in VN in with the actual humanoid model.
Fig.1. Model of wheelchaie in VN in various posisions.
2.2 The Control system The initial aim of the controller is to lift the wheelchair, and then stabilize it in its position. The wheelchair must be able to carry a range of weights (50-100 kg). The two back wheels of the wheelchair are fitted with torque motors, which should apply enough torque on the back wheels to lift the whole wheelchair upwards while moving a very small distance (forwards or backwards) from the original position. The torque must not be too high otherwise the wheelchair will flip over, or the wheels will slip, or it will lift up very quickly resulting in instability. The torque must not be too low, or else the wheelchair will move forward and not lift up. At the start of the lifting stage, a large torque results in a high angular velocity. The velocity should decrease as the wheelchair comes closer to the upright position, and once the wheelchair is upright the velocity must be zero. The better the velocity regulation is, the smoother the wheelchair lifting would be. The angle the wheelchair makes with the horizontal (ground) is 140, so by testing and with basic mathematical calculation the angle at which the
972 S. C. Gharooni, B. Awada and M. O. Tokhi
wheelchair is upright was found to be 50 (140 90 = 50). This is an important value because it is the controllers set point or target. The amount of initial torque required to be applied was found out to be around 100 N to 150 N depending on the weight on the chair. This value was found by experimental testing on the wheelchair model, where a manual slider control was connected to the rear torque motors and different torques were applied until a torque was found that was able to lift the wheelchair appropriately i.e. not too fast and not too slow. 2.3 Designing the Chair Rotation Control Rotating the chair depends on the wheelchair orientation angle. Therefore, the controller input was chosen to be the wheelchair angle and the controller output was chosen to be the chair orientation angle. To rotate the chair an orientation motor was fixed between the chair connecting rod and the axis connecting rod. In the first 0.5 seconds of the control the wheelchair has to rotate gradually to the back by 20, and then the control will shift to start rotating the wheelchair up. Once lifting takes place then the chair has to start rotating This rotation must coincide with the wheelchair lifting, so for different values of the wheelchair angle a chair angle was set, the values are shown in Fig. 2. Mass
Wheel Angle
140 130 120 110 100 90 80 70 60 50
Chair Angle -20 0 20 30 40 50 60 70 80 90
Chair rotation
Chair Angle
Wheel Angle
Wheelchair rotation
Fig. 2. Wheelchair and chair rotation during lifting up process.
Modeling and control of upright lifting wheelchair
973
2.4 Designing fuzzy logic controller The idea of fuzzy control is to build a fuzzy model of a human control expert who is capable of controlling the plant without thinking in terms of a mathematical model. The main advantage of the FLC approach is the possibility of implementing rules of thumb, experience, intuition, prediction and heuristics without the need for a mathematical model [7]. To design the fuzzy controller, membership functions (MFs) for both inputs and output must be drawn. The variables that are required by the controller as input are error and change of error of the wheelchair orientation. The change of error MFs were designed by observing the system and recording the maximum and minimum change of errors reached so that MFs can be designed. The MFs for error has a minimum of 50 and a maximum of 90. This is due to the reason that at the start the angle is 140, which means that the maximum positive error is 90. The torque requirement was found to be between 150 Nm and +150 Nm. 2.4.1 Fuzzy rule base A set of fuzzy if-then rules is contained in the fuzzy rule base, and the fuzzy inference engine uses these rules to determine a mapping from fuzzy sets in the input to fuzzy sets in the output universe based on fuzzy logic principles. It was found that the easiest way to design a rule-base was by deciding on where the OK has to be for the system to become stable or to reach the required target.
Table 1. The Error and change of error rules table
e
NM NS OK PS PM PL PVL
NVL PL PL PL PL PL PM PS
NL PL PL PL PL PM PS OK
NM PL PL PL PM PS OK NS
NS PL PL PM PS OK NS NM
NO PL PM PS OK NS NM NL
OK PM PS OK NS NM NL NL
PO PS OK NS NM NL NL NL
PS OK NS NM NL NL NL NL
The first OK was placed in the zero error, zero change of error cell (the dark shaded cell in Table 1). The other OKs were then placed diagonally above and below. As the OKs are in their position, the next decision that had to be taken was; to which side of the OK shall a negative or a positive torque be placed. The continuity rule was used to fill up the re-
974 S. C. Gharooni, B. Awada and M. O. Tokhi
maining cells. This means that after the PS, an OK will come then an NS, followed by an NM and etc., i.e. no jumps of torque output between the neighboring cells, ensuring that the control is smooth and stable. 2.4.2 Upright stability Although the above controller was designed for the lifting stage it can be used for the stabilization stage too. The rules designed for both controllers are the same, so the rule-base does not have to be changed. The same MFs can be used, and the only change would be to the input signal, i.e. when the final stabilization starts to take place, the error would be very small. Therefore a gain can be placed to step it up, so that the controller can give a finer and more precise output. 2.4.3 Implementing controller Following the controller design, Simulink design, and tuning of the parameters (output torque gain, the change of error multiplication graph, and the error gain), the controller was tested with a 70 kg weight, and for some disturbance forces applied to the wheelchair. The final Simulink block diagram incorporating the controllers, the plant, the chair rotation control, and the controller switching action, is shown in Fig. 3.
Fig. 3. The error/change of error control in simulink.
Modeling and control of upright lifting wheelchair
975
3 Results
To test the controller a force was placed on the seat in the y-direction i.e. as if someone is pushing the chair from the back or from the front. This experiment was performed to test if the wheelchair is pushed by mistake from the most dangerous area (the top of the chair) and whether the wheelchair will stay upright and stable. The position where the disturbance force was applied and wheelchair trajectory transient to upright are shown is shown in Fig. 4. It shall be noted that although the figure shows that the forces are in the same direction, they are not in practice, as this is the total modulus of force. The first impulse is positive, the second is negative, and the third is positive. All the forces were applied for a period of 0.12 s.
|F| (N) vs. time (s)
Rz (deg) vs. time (s)
600
140 130
500
120
400
110 100
300
90
200
80 70 60
100
50
-100 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
40 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Fig. 4. The disturbance impulse force and the wheelchair angle.
Fig. 5 (left) shows the angular velocity of the wheelchair. As noted, the velocity at the start increases and then decreases with few oscillations in between, hence a smooth lifting control. The wheelchair lost its speed and stopped oscillating when a 200 N or a 500 N force was applied.
Wz (deg/s) vs. time (s)
Tz (N m) vs. time (s)
100
120 100
0
80
-100
60 40
-200
20 0 -20
-300
-400
-40 -60
-500
-80
-600 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
-100 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Fig. 5. The wheelchair angular velocity and torque (y-axis) against time (xaxis).
976 S. C. Gharooni, B. Awada and M. O. Tokhi
It can also be noticed from the torque graph (Fig. 5 right), that when the controller reaches the set-point, the torque oscillates slowly between 4N and 5N, i.e. the oscillation frequency is not high, which means that the torque moves smoothly from 4N to 5N and vice versa just to keep the wheelchair upright, which is normal and causes no vibration.
4 Conclusion
The controller was tested for the following functions: Lifting the wheelchair upwards in a stable smooth manner Stabilizing the wheelchair in an upright position and stopping it moving. Lifting weights in the range of 50 to 100 kg. Disturbance adaptation, i.e. if a force is applied on the wheelchair, it will not deviate significantly from the set point, and it will be able to re-stabilize itself. As demonstrated, the controller reduces the angle from 140 until it reaches the set point (50).
References
1. Wellman P, Krovi W, Kuma V, Harwin W, (1995) Design of a Wheelchair .with Legs for People with Motor Disabilities, IEEE Transaction on Rehabilitation Engineering, Vol. 3, pp. 343-353. Johnson and Johnson ( 2004) website: [Link] (Date of last update: March 2004) Cooper RA, Boninger ML, Cooper R, Dobson AR, Schmeler M, Kessler J, Fitzgerald SG. (2003) Technical Perspectives: Use of the Independence 3000 IBOT Transporter at Home and in the Community, Journal of Spinal Cord Medicine, Vol. 26, No. 1, pp. 79-85. MSC SOFTWARE. Stunt Man, free Visual Nastran demo downloaded from website [Link] Ross, T.J., 1995, Fuzzy Logic with Engineering Applications, (McGrawHill, New York). The Math works, Version 6.5.0.1890913a Release 13, June 19, 2002, Matlab. Lee, C. C. (1990a). Fuzzy Logic in control systems: Fuzzy Logic Controller Part I. IEEE Transcations on Systems, Man and Cybernetics, 20(2), pp. 404-18.
2. 3.
4. 5. 6. 7.