The History Of Grimblebot

Published Date: 02 Nov 2017

GRIMBLEBOT

This report will discuss and evaluate the performance and construction process of the grimblebot. It will give an explanation of how the Grimblebot circuitry works, including a few simulation results to demonstrate key functions. It will outline key issues relating to the placement of parts and the design of the PCBs, explaining why the components are located where they are. It will also give an understanding of difficulties encountered in the project, including reasons and solutions. Finally, a summary of the findings, including recommendations to improve the Grimblebot and modifications that could be required to manufacture the Grimblebot in large volumes will be discussed.

SELF BALANCING ROBOT

Contents

9 APPENDIX.....................................................................................................................................14

1 INTRODUCTION

1.1 Overview and context

The Grimblebot is a self-balancing robot that stands on only two wheels. It behaves like an inverted pendulum with the weight of the body above the pivot point, and this makes the system is very unstable. Therefore the device would need to constantly move its base (the wheels) of the system to achieve stability. These wheels which are connected by a gear system to a motor move to the direction the robot intends to fall allowing the robot to re-establish an upright position. In order to achieve balance, the device would need to measure the angle of inclination of the robot with respect to the vertical and drive the motor in such a way to ensure that the robot remains balanced. It follows therefore that the following would be put into consideration: a means of measuring the position (a transducer) of the robot, executing the negative feedback control algorithm controlling the motor (a microprocessor), amplifying the signal from the microprocessor and driving the motors (motor drivers) attached to the wheels.

Fig 1.1: A simple Schematic drawing of the Grimblebot

When held in a vertical position and switched on, the motor starts and moves in the direction which the robot intends to fall, so that the motor torque about the centre of gravity which is higher than the motor makes the robot balanced. The Grimblebot is then tested on different surfaces to see how long it balances on each of them, and conclusions are drawn with regards to its functionality from the results obtained.

1.2 Objectives

The purpose of this project was to build and evaluate the Grimblebot in order to analyse and evaluate its performance on different surfaces as well as examine possible modifications that could be implemented in the construction process that would make it suitable if it were to be considered for mass production.

The report will explain how the Grimblebot works, how well it works under different conditions and any issues that were experienced during assembly and testing. It will also show results of simple simulations and relevant results obtained from the analysis of its functionalities as well as conclusions drawn from the results obtained.

2 IMPLEMENTATION

This section explains the procedures taken in the construction of keys parts of the Grimblebot.

2.1 Construction of the Ultrasonic Transmitter/Receiver Board

The ultrasonic transmitter/receiver board (daughter board) takes a 50 KHz signal at 3.3v from the mother board and produces an amplified ultrasonic signal. This signal which is generated by the microprocessor on the mother board would then be transmitted by the transmitting sensor, and concurrently received back to the board through the receiving sensor. This would then be amplified and using a comparator, sent back to the motherboard as a digital signal between 0v and about 3.3v.

A schematic of the board was implemented using the OrCad software package, (see Appendix 1 for schematic). This schematic included the transmitter and receiver circuitry. It was important to check that the output was consistent with what was expected. The circuitry was simulated and the output signals observed. The voltage trace was accurate, and showed that the output voltage to the transmitter was a square wave that oscillated from +7.2v to -7.2v, (See Section 3.1 for Voltage Trace). The reason for manipulating the signal was that it essentially doubled the voltage, producing more power behind the transmitter.

Once the system behaved as expected, the circuit was invoked with OrCad Layout which allows the components to be placed in suitable locations in a digital PCB. The digital PCB board can now be created with the print off of the layout from OrCad by a process called etching. After the etching process has been completed, the board is then drilled at desired positions and its components fitted and soldered.

2.2 The motherboard

In the construction of the motherboard, there are four stages involved: populating the board with surface mount components, through hole components soldered onto the board, soldering of the battery packs in place, and lastly mounting the gearbox and wheels onto it.

The motherboard was secured with PCB holders and held with a masking tape, then a stencil similar to the shape of the motherboard which also had layers where the solder was to be placed on the board was mounted over the motherboard. A toothpaste sized amount of solder was then put on the stencil and spread through it using a flexible palate. The different surface mount components are then fitted to the motherboard and baked in the oven.

2.3 The Gearbox

The gearbox is to be attached to the front of the motherboard. A 12mm M3 bolt is given with an M3 tooth lock wash and nut on each side of the motor. The washer and nut are used on the back of the motherboard to secure the gearbox in place using the bolt inserted from the front of the motherboard.

3. RESULTS AND FINDINGS

This section gives an analysis of how the Grimblebot works, along with important factors to consider when constructing the board. It also shows results of simple simulations and relevant results obtained from the analysis of its functionalities as well as conclusions drawn from these results obtained.

3.1 How the Grimblebot Works

To start the robot, it needs to be set up in an upright vertical position before the switch is turned on so that the signals being transmitted by the ultra sensors can bounce off the surface on which the device rests and be received by the receiver ultra sensor. In the first few seconds signals are sent by the microprocessor via the ultrasonic transmitter/receiver to obtain a reference distance between itself and the ground. Its motor is turned on and its wheels then begin to move, as it attempts to maintain this reference distance that it ‘perceives’ as its balance point. The beige rheostat on the motherboard may need to be calibrated in order to fine tune the balance of the device.

3.2 Theory behind its self-balancing mechanism

The Grimblebot is powered by 6 batteries supplying a total of 7.2 volts. Using voltage regulators, it also has power rails to supply 5.0 volts and 3.3 volts. These voltage regulators provide high quality, clean and well regulated power sources for various components. The microprocessor functions under a 3.3 volt supply, and generates 12 square wave pulses at 50 KHz at 3.3 volts. It sends this signal to the daughter board which has the ultrasonic transmitter and receiver.

This daughter board takes this signal through a transistor working as a switch and allows 7.2 volts from the main power supply to be propagated through this board. The output response of which is shown in Figure 3.2.1 This signal is then boosted to give a potential difference across the ultrasonic transmitter of 14.4 volts ranging from -7.2 volts to +7.2 volts. The output response is shown in Figure 3.2.2

Fig 3.2.1 Input response from 0 to 3.3 volts (solid trace). Output response from 0 to 7.2 volts (broken trace).

Fig 3.2.2 Input response from 0 to 3.3 volts (solid trace). Output response from -7.2 to +7.2 volts (broken trace).

The output ultrasonic wave is then reflected off the ground, and returns to Grimblebot. This signal is then picked up by the ultrasonic receiver. This signal has amplitude of only 60mv; therefore this would need to be amplified. The set up uses an operational amplifier to amplify the signal to the power rail supply of 3.3 volts. This required an op-amp with a gain of 50. It then uses another operational amplifier as a comparator to switch the signal between 0 volts and about 3 volts, (the above is demonstrated by Figure 3.2.3). This creates a digital waveform to be sent back to the motherboard. It is important to note that the signal will always have the same frequency as it did on being created; that of 50 KHz.

Fig 3.2.3 Signal is amplified to between 0 and 3 volts. It is then passed through a comparator which switches the signal either off (0volts) or on (3volts).

An important point to note the op-amp acting as an amplifier also functions as a first order, low-pass filter with a cut-off frequency of 70 KHz (the response of which can be seen in Figure 3.2.4). This is necessary to filter out any high frequency noise that the receiver might pick up.

Fig 3.2.4 Bode Diagram of Op-Amp with cut-off Frequency of about 70 KHz.

When the microprocessor sent out its signal, it also started a timer. When the signal was returned via the ultrasonic receiver, the timer was stopped. Knowing the frequency of the signal, and the time taken to propagate, the distance between the Grimblebot and the ground can be calculated.

Using this information the microprocessor can now drive the motor in the correct direction and amplitude using a Proportional-Integral-Differential algorithm. However, the microprocessor cannot directly drive the motor as it does not have the current delivering capability to power it; furthermore it can only produce a digital signal which is not desirable, the speed of the motor needs to be varied.

Since the output from the microprocessor is not sufficient enough to drive the amplifier which requires an analogue signal as its input, the device uses an H-bridge to drive the DC motor in a bi-directional with variable speed. It works by switching transistor pairs on or off. The pairs of transistors are diagonally opposed to each other. If the high side left and low side right switches are turned on, current passes as shown in Figure 3.2.5

http://www.pt-boat.com/electric/images/hbridge.gif

Fig 3.2.5 H-Bridge schematic showing current flow

This provides a potential difference between the motor terminals, making the motor rotate. If the other pair of transistors are switched on (and consequently the other pair was switched off), the motor would drive in the opposite direction. The H-bridge transistors are switched using the 3.3 volt power supply from the microprocessor, and this allows the 7.2 volt power supply to flow through supplying enough power to drive the motors.

3.3 Testing the Grimblebot using an oscilloscope

The hardware of the robot was tested to debug it using an oscilloscope. This device gives an overview of the voltages in a circuitry with respect to time. In testing the Grimblebot, the voltage flowing through the components closest to the power supply was 8.3 volts; which was within the expected range (7.2v - 9v).

Various pins from the daughter board connection were also tested. The main power supply received 7.22 volts and the secondary power supply received 3.43 volts. The pin providing the signal from the microprocessor received a 3.6 volts signal at 50 KHz as shown in the figure below.

J:\TEK0001.BMP

Fig 3.3.1 3.6 volt signal from Microprocessor at a frequency of 50 KHz

The Grimblebot was put in an upright vertical position as it would when put on its own to balance, but without connecting the motors. Instead a 10k resistor was connected to the motherboard and the MP to US pins and the US to MP pins were tested to evaluate the input and output responses of the ultrasonic transducers. There was a change of time between the transmitted signal and the reflected signal as can be seen in figure 3.3.2. From the diagram, the first signal on top is the input signal, while the lower one ( second one) is the received output signal.

J:\TEK0003.BMP

Figure 3.3.2 Input and output signals of the device showing time delays between them.

3.4 Performance evaluation/testing on various surfaces

Surface type

Time stayed balanced

Approximate displacement values

Comments

carpet

Just over 15 seconds

30 cm

The device stayed relatively stationary in the first 5 seconds. Then it moved forward and eventually fell in its side

Tiled floor outside G26

30 seconds

5cm

On turning the power of the device on, it made a few oscillations but managed to balance for about 30 seconds.

Curved lab chair surface

Over 34 minutes

N/a

This was the most promising result; The device managed to balance itself, only oscillating backwards and forwards by very nearly unnoticeable oscillations. It eventually fell off the table probably due to the battery running low.

Table showing performance of the Grimblebot on different surfaces

After the grimblebot had been debugged, an evaluation of the device’s performance was done to see how it responded with regards to surfaces. The table shown above summarizes the result of this analysis.

3.4.1 Deductions

From the evaluation of the Grimblebot on different surfaces, it will be wise to infer that there is a reasonable explanation as to the different balancing times on these surfaces. The reason for variations in time can be explained thus: reflection of sounds and signals occur differently on different surfaces so also the amount of these sounds and signals received by the receiver in the ultrasonic sensor after transmission.

On the rug surface the Grimblebot did not balance for long as it did on the tiled floor or the chair surface due to the transmitted signals not bouncing back at neatly as they would on the other two surfaces. This comes from the idea that on carpeted surfaces less ultrasonic waves will be received, because the carpet will absorb some This made the receiver not to get back the desired response the device needed to adequately measure the distance to ground and balance using it as reference.

On the curved chair surface, the Grimblebot balanced for very longer due to very good reception of the transmitted signals, and due to the slightly curved edges of the chair which made the device find a suitable centre of gravity for it to balance.

Other issues that could be responsible for the quality of the devices functionality are: People walking around it causing the ground to vibrate, external disturbances of the surfaces, the vertical orientation of the robot as it is switched on, the charge of the battery used, and the beige rheostat calibration.

4 ISSUES RELATING TO PLACEMENT OF PARTS

Placement of components and parts on the Grimblebot was an important aspect to consider. It is clear that, in the real world, connections, wires and tracks on a PCB board all have an associated resistance, capacitance and/or inductance to them. These aspects, small as they are, could have an overall effect on the performance of the Grimblebot. Therefore it was vital to reduce the number of these connections, make distances between components as small as possible and reduce and if possible eliminate any jumpers on the board.

To minimise the occurrence of jumpers, the components related to any of the test pins were placed close to the test pin. However, the daughterboard constructed in this device had three jumpers. After the parts had been placed correctly, the routing process provided a board as can be seen in the appendix.

The result was observed when the board was tested, as the variance between the simulated responses and the actual responses was very small.

The same concept could be applied when looking at the mother board. The component positions were already pre-determined. However, parts were placed in such a way as to be as close as possible to the components that they related to. For example, the H-Bridge IC was placed very close to the motor; or the microprocessor was placed in the centre of the board, with components surrounding it.

Another key issue was found in the routing process. Test pins J4 and J8 had been locked to help with the routing process of the board. If these pins had been unlocked and moved, there would not have been complications in the routing process and the final complication of the board

4.1 Things that can go Wrong

This section examines any issues that were encountered during construction of the Grimblebot; and any potential problems or issues that could occur with system.

Issues during Construction and Possible Solutions

During the construction process a few minor mishaps or delays had occurred, due to either not reading the brief carefully or simply from a lack of experience with the tools used. Below were the most relatively significant issues;

Main LED: The motherboard used a self-flashing LED that would switch on when the power was switched on. It did not flash initially, however this did not affect the performance of the Grimblebot. Further examination showed that this specific LED was damaged and therefore needed replacing.

Jumpers: The device had three jumpers at the end of th3e routing process. Although this did not affect its electronic correctness, it created more possibility of short circuit if it was exposed.

Issues discovered while Testing and Possible Solutions

While testing the Grimblebot and seeing it operating first hand a number of potential issues became apparent.

Safety features: Evidently, this device will be subject to same toppling over from now and again. Therefore it might be worthwhile investing in stronger parts. The ultrasonic transmitter kept on having to be straightened every time the device toppled over

Lubrication: The motor used was a simple one, which served its purpose. However the quality of the motor and gearbox setup could be improved by adding more lubrication to the moving parts. This will allow less energy lost due to friction and may even fill in the extra spaces between the teeth of the gears.

Component error: During the project, some of the components provided did not measure accurately to the specified value required. For example in place of a 1k ohm resistor, a 1.2k ohm resistor was used and this added to other such small variations in other components could hamper optimum performance of the device.

Noise: The effect of noise on the Grimblebot creates disturbance to the sytem feedback control mechanism and could affect its performance.

Heat: Heating of the components or of the battery could lead to the device not functioning properly as most electrical components are heat dependant and can be adversely affected by small temperature changes to them.

Inductance: The inductance of the capacitors used should be such that they can adequately control the energy flow in the system.

Gearbox: When the gears in the kit given were linked together it was noticed that there were some extra spaces between the teeth. This made them a little bit shaky, allowing the wheels to turn about a degree or more in either direction. This might affect the level of performance when considering that this is a delicate balancing system. A possible solution would be to invest in better quality gears.

Batteries: Each battery type available will output a slightly different current to the device and this may need to be factored in. If the Grimblebot receives a much smaller current then it perceives, then the wheels will react too slowly; similarly, if it is supplied with too much current the wheels will react much faster. This therefore means that with every change of the batteries used, the beige rheostat would need to be caliberated

Other possible factors that could be put into consideration could be the type of microprocessor used the quality of soldering connections on the board and so on.

RECOMMENDATIONS AND MODIFICATION

From research it is easy to observe different procedures used in making self-balancing robots. It is interesting to observe the use of the devices below which could be useful in achieving perfect stability of these robots.

Gyroscope

A device consisting of a spinning mass, typically a disk or wheel, mounted on a base so that its axis can turn freely in one or more directions and thereby maintain its orientation regardless of any movement of the base [2]. The gyroscope can also be integrated to provide accurate dynamic tilt information, but the integration tends to drift over time [6]. Therefore if this device had been used in this project, the grimblebot should balance indefinitely and would improve the production of the self balancing robot. The gyroscope is not usually used alone as it cannot measure acceleration of the grimblebot therefore a device known as an accelerometer is used. The gyroscope can also be found in video camera to balance handling and also to balance the tail of an helicopter [1].

Accelerometer

The accelerometer basically measures the acceleration of an object and is very useful in the construction of a craft and is also used in guided missiles. The accelerometer if used in this experiment could provide more accurate static tilt information when the robot is accelerating. Therefore, the effect of different substances absorbing signals transmitted will not be a problem. The accelerator can also be used to correct drift.

Therefore if the gyroscope and the accelerator has been used combined with complimentary filters stability of the grimblebot would have been much improved.

Complimentary Filters

The idea of the complimentary filter will be to combine the output of the accelerometer and the gyroscope to obtain a good estimate of the orientation thus compensating for the drift of the gyroscope and for the slow dynamics of the accelerometer [3].

Rough edged wheels:

Using a wheel with rough edges could improve the stability of the device when it balances; reducing the frequency of times the robot would quickly run off when turned on.

Programming: It could be possible to allow the user to re-program the PIC on the mother board to change the response of the motor. This would require another part to be fixed onto the board and possibly a software package to be included with the product.

6 CONCLUSION

From the results and analysis of data in this report so far, it can be seen that the objectives of the report has been met: A description of the building and coupling process of the device was discussion as well as the difficulties encountered during this process. Tests were carried out to test the functionality of the device, and its overall performance on different surface types was analysed, and from the deductions drawn, it can be observed that the robot exhibited a better performance when tested on a curved lab chair surface (wooden). This when compared to results from the tiled floor outside the lab, and the carpet surface in the lab, showed that hindrances such as loss of signals transmitted, as well as disturbances of surfaces by external factors could affect the functionality of the robot causing it not to balance for long.

The report also highlights problems encountered during the construction process and possible solutions as well as modifications that could be made to the manufacture of the device to enhance its performance and functionality. Among these factors include the use of a gyroscope, accelerometer, and complimentary filters, etc. A major factor which was not included in the report could be the soldering system used during the construction process especially if the device were to be considered for mass production. A better option might be a consideration of surface mounting techniques on the daughter board.

7 ACKNOWLEDGEMENT

I would like to acknowledge my project partner Mary Agarry for her total support and commitment to the successful completion of the project. I would also want to recognise the support of Mr John the ever effective helper and technician in G26, and finally Dr Ben Potter for his patience and invaluable assistance throughout the project.

8 REFERENCES

[1] Building a balancing scooter, http://tlb.org/scooter.html

[2] Gyroscope definition, http://www.instructables.com/id/Balancing Robot/

[3] David P Anderson, nbot balancing robot, http://www.geology.smu.edu/~dpa-

www./robo/balance/inertia.pdf

[4] David PAnderson, nbot balancing robot, http://www.geology.smu.edu/~dpa-

www/robo/nbot/

[5] John Drain, Brushed Motor Speed Controller Designs http://www.pt-boat.com/electric/speedcontroller.html

All others: Shirsavar and Hallworth,(2009) Grimblebot Project Manual (2.0) For Engineering Applications, University of Reading.