Micromouse Design Specifications
Prepared for:
Dr. Dan Lewis
Santa Clara University
by:
Jonathan Herbst
Philip Livengood
Kevin Wollenweber
June 2, 1998
Abstract
“Micromouse Design Specifications”
by: Jonathan Herbst, Philip Livengood, and Kevin Wollenweber
This project is designed with the intent of entering it in the 1998 IEEE
Micromouse competition. This micromouse will need to be small, completely
autonomous, and intelligent. For the competition, the micromouse is allowed two passes
in the maze. The first pass is used to completely learn and map out the maze in an attempt
to find the center, or goal of the maze. Then the second pass should find the shortest path
to the center in the shortest time.
With these goals in mind, we have designed a small, square shaped Micromouse.
There will be eight infrared sensors, two on each corner of the mouse which are used for
guidance. There are two wheels driven by stepper motors with the remainder of the
weight balanced by two castor wheels. These wheels will be driven in a tank-like fashion
using the infrared sensors to maintain its alignment. Our design also includes two power
sources, one to drive the motors and one for the remainder of our components.
The intelligence of the micromouse is based on its software. The goal of the first
pass is to see as much of the maze as possible in order to create a map of the maze. In
turn the second pass should take this map and determine the quickest path to the center.
Abstract .......................................................................................................................... 2
Introduction .................................................................................................................... 4
Design Interaction........................................................................................................... 5
Design............................................................................................................................. 6
Motion.............................................................................................................................................. 6
Sensors............................................................................................................................................. 8
Power ............................................................................................................................................... 9
Control ............................................................................................................................................. 9
Software ......................................................................................................................................... 10
Conclusions................................................................................................................... 13
Appendix A: Micro Mouse Contest Rules...................................................................... 15
Appendix B: Software Flowchart.................................................................................. 18
Appendix C: Prototype Motor Driver Assembly Code................................................... 19
Appendix D: Micromouse Software .............................................................................. 20
Appendix E: Dijkstra’s Shortest Path Algorithm........................................................... 21
Introduction
The Micromouse competition is a national competition that involves building an
autonomous robot which can traverse a maze and intelligently calculate the fastest route
to the goal of the maze then run to it. We have decided to build a Micromouse for our
senior project so that it conforms to the official IEEE Micromouse contest rules (see
Appendix A) with the intent of entering it in this competition after our design project is
finished. Our goal is to have our mouse collect very accurate data about the maze so that
even if it is not the fastest mouse, it has a good chance of finding the best route through
the maze. To achieve this, we have decided to focus on the sensors and the software that
controls them. This report will discuss the different design ideas we have come up with
and what direction we are planning on going with our Micromouse.
Design Interaction
For our Micromouse, we decided to design our robot so that it meets the IEEE
Micromouse Contest (see Appendix A) rules. To make the project easier to deal with we
broke it down into five sections: motion, sensors, power, control, and software. The
control section, consisting of the microprocessor and controlling hardware, receives
information that is collected from the sensors section which acts as our data collection of
the maze environment. Control is then responsible for sending information to the motion
control section that drives motors and wheels. Power is simply responsible for powering
the components and motors. The brains of our mouse will be the software that is run by
the microcontroller.
Software
Motion
Control
Power
FIGURE 1: DESIGN INTERACTION
Sensors
Design
Motion
During our design analysis, we were very concerned with the shape of our mouse
and whether its shape would inhibit its movements through the maze. Our first concern
was whether a square shaped mouse would catch itself on the corners of the maze, so our
first designs focused on a circular or octagonal shaped mouse. Although these designs
were never tested, we felt that they would not work for the types of sensors we wanted to
use. If we had eight infrared (IR) sensors (one on each edge), there would be no corners
to mount them on. In the case of the octagonal mouse, the IR sensors may cross each
other and somehow get disrupted or trigger another sensor’s detector. This is the main
reason we decided on a square Micromouse. We can place an IR detector/emitter pair on
each corner (see Figure 2) of the mouse so that we have sensors at each edge. The main
Castor Wheel
Drive Wheels
Sensor pair
FIGURE 2: MICROMOUSE (TOP VIEW)
drawback to the square mouse is that it will have a wider turning radius and may catch on
corners, but we feel that if we focus on using the sensors to make sure we are turning 90°
each time, we have eliminated much of this problem. In using the square design, we will
have to make sure our mouse does not become too large or it may have problems turning
tightly in corners.
One major area of focus has been in the area of wheel structure. One way to make
sure our mouse was turning 90° was to simply turn the wheelbase 90° and we would not
have to troubleshoot our turns as much because we knew how much it was turning. A
problem with this design was to turn the wheelbase without friction so that we could
ensure the 90° turn. To solve this, we would have needed to attach a solenoid to the
wheelbase and lift it into the mouse while the mouse was supported by its castor wheels.
We would then turn the wheelbase and put it back down. This would have created a
mouse that would almost guarantee a 90° turn. We were also concerned with the
possibility that the mouse would somehow end up in a non-90 degree angle and we
would not be able to correct it. With this design, small increments would be very
difficult. So, we would have to try and use the wall to realign our robot. We were very
worried that this type of trouble shooting was not very reliable. This is why we decided
on a different type of locomotion.
Our solution is a mouse driven by two stepper motors which will control the two
main wheels. Stepper motors are the obvious choice for this project since they can be
controlled to turn a certain distance with each clock pulse they are given. By counting
the number of pulses, the mouse can be controlled to go a certain distance and its position
in the maze can be calculated easily. With two wheels driven by stepper motors, and two
castor wheels to help support it (see Figure 2), our mouse will move much like a tank. To
go forward, we drive both wheels at the same rate. To turn, we simply step the wheels at
the same rate, but in the opposite direction. We can also make adjustments to the mouse’s
position by giving small pulses to the stepper motors. We felt that, although this
technique is much more error prone than the first, it would be much simpler. In keeping
our design simpler, we could use our extra resources to make sure our mouse is turning
90° and is centered in each hallway.
Sensors
The mouse will have eight IR emitter/detector pairs (see Figure 2) that will keep
track of the presence of either walls or doors, and will be mounted on each corner of the
square. To gather information about the structure of the maze, we will pulse the IR
emitters. If this pulse is received in the detector, then the light has bounced off
something and there is a wall in front of it. If there is no signal received, then there is a
door in the direction of the sensor. In early tests, we have found that enough ambient
light in the room can disrupt the reception of the IR signal, so we have decided to pulse
the IR emitters at a certain frequency, then filter out that frequency on the receiver side to
help solve this. This way, we can filter out the noise from the ambient light and send
powerful pulses for a short duration to make receiving the signals easier. This data will
then be sent to the microprocessor and software algorithms will decide which direction
the mouse should move.
Power
For our power control, we will be using two different sources: one supply for the
stepper motors and one supply for the remainder of the components. The reason we are
using two different sources stems from the problem that the stepper motors draw a large
amount of current when they start their motion. This current draw could cause our
microprocessor to be reset, in turn halting and restarting the execution of our code.
For both of our power supplies, we will use a simple array of batteries. Since all
of our components require 5VDC, we are going to use an array of batteries producing
more voltage than we need and then stepping this voltage down to 5VDC. This will
allow our batteries to last longer. The stepper motor drivers require a maximum of 24V
to run. We have tested them and have decided that using 16 rechargeable AA batteries
will give us sufficient power to run the desired time.
Control
The Micromouse controller must take directions from software and control the
mouse’s peripherals. It must control the motors and the sensors while executing the
microcode that it fetches from program memory. We have designed our initial mouse
controller using an 8051 microcontroller. The chip comes with some memory on board,
both RAM and EMPROM, but it is not enough for this project. The code for the
algorithms used and the data structures that store the maze data take up a lot of memory,
so we are going to add 32K of external EPROM and 64K of external RAM so we are not
limited in our program size. The 8051 has four input and output ports (two of which are
used for memory), so we need to add a peripheral controller for a few extra output ports.
We will use one port to strobe the IR LED’s, one to detect the IR signal, one to control
the motors, and one to give ourselves some room to add features (debugging lights and
switches) we need during additional testing (see Figure 3).
A
PORT A
B
PORT B
C
D
PORT C
DEBUG LED’s
MOTOR
DRIVERS
DIP SWITCH
PORT D
ADC
8051
ADDRESS/DATA BUS
STATIC
RAM
ROM
FIGURE 3: HARDWARE BLOCK DIAGRAM
Software
The software (see Appendix D) is broken down into three sections to make
it easier to build and test. The first section is the component control, which deals with
control of the sensors and motors discussed earlier. This code is written in assembly
language. Assembly language is much simpler and faster when interfacing with
SENSORS
hardware directly, and will save space in program memory for the more complex
algorithms. The second section is the first pass algorithm which is implemented in C.
Assembly language is much faster since it directly manipulates the hardware and can be
optimized, but C is only slightly slower when it is compiled into assembly language and
made programming the algorithms much easier to understand and debug. It may cost us
time, but it is on the order of microseconds and will not effect the overall outcome of the
race.
The basic principle behind the first pass algorithm is to traverse the maze and
create an undirected graph in memory that will be used later by the shortest pass
algorithm. The third section will be the shortest path algorithm that will look for not only
the shortest distance to the destination point of the maze, but also for the path that will
take the shortest amount of time to reach the goal. Will have chosen to use Dijkstra's
shortest path algorithm (Appendix E) These two are not necessarily the same path since it
takes the mouse much longer to turn a corner and start again than to continue going
straight.
During the first pass, the mouse will follow these basic rules at each square:
-Try to go right
-If it can’t go right or already visited the square, go straight
-If it can’t go straight or already visited the square, go left
-If it can’t go left or already visited the square, turn around
As the mouse moves through the maze, decisions must be made at each square.
Following our basic rules listed above, if a turn must be made and we have not visited
this door before, then this square is called a node and it must be added to our undirected
graph. Also, if there are more doors at a particular square that we still need to visit, then
we must store the coordinate of this square in a stack data structure in order to return to
this location again. Each time the mouse finds a dead end and must turn around, then a
coordinate is taken off the stack data structure and the shortest path algorithm is used to
determine how the mouse is to move from its current location to the coordinate which
was just popped off the stack.
The graph that is created from the first pass algorithm, stores the coordinates of
each node, the distance between connecting nodes in the maze, a count of the number of
nodes in the graph, and the direction that the mouse is facing within the maze. The mouse
continues to search the maze until it has visited every square in the maze and there are no
more coordinates in the stack data structure mentioned above. Having done this, the first
pass algorithm is complete. Figure 4 illustrates an example of the first pass algorithm and
the data collected within the data structures listed above. In the figure, the dotted lines
indicate the movement of the mouse, the circles indicate the locations of the nodes and
the blackened circle indicates the current location of the mouse. Once the first pass
algorithm has completed, the mouse will then be placed at the starting point of the maze
and the shortest path algorithm will then determine the fastest route from the start to the
destination square. This completes the micromouse simulation.
Graph Data Structure
Facing: North
Node Count: 4
Nodes:
(0,0) (0,1) (2,1) (2,2)
Edges: (0,0)
0
1
0
0
(0,1)
1
0
2
0
(2,1)
0
2
0
1
(2,2) 0
0
1
0
FIGURE 4: SAMPLE MAZE AND GRAPH DATA STRUCTURES
Conclusions
This was a large design project and there were a lot of aspects of the project that
we were not too familiar with, but we learned a lot. This project was perfect for a senior
design project because it involved knowledge of electronics, computers, and mechanical
design. So far the project is about 80% completed. We finished the structural design of
the mouse and the software to control the mouse. Once the sensors are functional, we can
begin assembling, testing, and debugging the system as a whole. We think that one of the
main obstacles we had was obtaining donated parts. The microcontroller that we had
donated by Philips Semiconductor was about 3 months late, which forced us to do our
development with the 87752. This type of development was difficult since we could only
test small portions of the mouse and not integrate parts together. Another obstacle we ran
into was a blown stepper motor driver. Our motors and drivers were donated by Oriental
Motors from discontinued stock. When one of the motors blew, there was no way to
replace it. We are still in the process of resolving this problem. Although the project is
not totally complete, we consider it a complete success for the time we had to finish it.
We would, however, do a few things differently if we were to start over.
One of the main things we would change about the way we went about this
project would be the development board. We did not buy a development board because
we decided to make one ourselves. We lost days of development time erasing and
reprogramming the EPROM. Development boards can be purchased specifically to
download test code into RAM, which is instantaneous and can be erased simply by
resetting the system. Erasing the type of EPROM we used involved placing the chip
under UV light for 20 minutes every time a different version of software is to be tested.
Another recommendation would be to spend more time on the sensors. The idea
of how the sensors are to work is fairly simple to grasp, but they can be difficult for
anyone who has little experience with them. Experiment with different ways of pulsing
the emitter so that you can send a powerful enough signal, at a certain frequency, that can
be detected by the emitter. These were the major obstacles we faced that have to be
resolved before this project can be completed.
Appendix A: Micro Mouse Contest Rules
(Adapted from IEEE/APEC Rules)
I. Specifications for the maze
1. The maze shall comprise 16 x 16 multiples of an 18 cm x 18 cm unit square. The
walls constituting the maze shall be 5 cm high and 1.2 cm thick. Passageways
between the walls shall be 16.8 cm wide. The outside wall shall enclose the entire
maze.
2. The sides of the maze shall be white, and the top of the walls shall be red. The floor
of the maze shall be made of wood and finished with a non-gloss black paint. The
coating on the top and sides of the walls shall be selected to reflect infrared light and
the coating on the floor shall absorb it.
3. The start of the maze shall be located at one of the four corners. The starting square
shall have walls on three sides. The starting square orientation shall be such that when
the open wall is to the "north", outside maze walls are on the "west" and "south". At
the center of the maze shall be a large opening which is composed of 4 unit squares.
This central square shall be the destination. A red post, 20 cm high, and 2.5 cm on
each side, may be placed at the center of the large destination square if requested by
the handler.
4. Small square posts, each 1.2 cm x 1.2 cm x 5 cm high, at the four corners of each unit
are called lattice points. The maze shall be constituted such that there is at least one
wall touching each lattice point, except for the destination square.
5. The dimensions of the maze shall be accurate to within 5% or 2 cm, whichever is less.
Assembly joints on the maze floor shall not involve steps greater than 0.5 mm. The
change of slope at an assembly joint shall not be greater than 4. Gaps between the
walls of adjacent squares shall not be greater than 1 mm.
II. Specifications for the Micro Mouse
1. A Micro Mouse shall be self-contained. It shall not use an energy source employing a
combustion process.
2. The length and width of a Micro Mouse shall be restricted to a square region of 25 cm
x 25 cm. The dimensions of a Micro Mouse which changes its geometry during a run
shall never be greater than 25 cm x 25 cm. The height of a Micro Mouse is
unrestricted.
3. A Micro Mouse shall not leave anything behind while negotiating the maze.
4. A Micro Mouse shall not jump over, climb, scratch, damage, or destroy the walls of
the maze.
III. Rules for the Contest
The basic function of a Micro Mouse is to travel from the start square to the destination
square. This is called a run. The time it takes is called the run time. Traveling from the
destination square back to the start square is not considered a run. The total time from the
first activation of the Micro Mouse until the start of each run is also measured. This is
called the maze time. If a mouse requires manual assistance at any time during the contest
it is considered touched. By using these three parameters the scoring of the contest is
designed to reward speed, efficiency of maze solving, and self-reliance of the Micro
Mouse.
1. The scoring of a Micro Mouse shall be done by computing a handicapped time for
each run. This shall be calculated by adding the time for each run to 30 of the maze
time associated with that run and subtracting a 10 second bonus if the Micro Mouse
has not been touched yet. For example assume a Micro Mouse, after being on the
maze for 4 minutes without being touched, starts a run which takes 20 seconds; the
run will have a handicapped time of: 20+4300-10=18seconds The run with the fastest
handicapped time for each Micro Mouse shall be the official time of that Micro
Mouse.
2. Each contesting Micro Mouse shall be subject to a time limit of 15 minutes on the
maze. Within this time limit, the Micro Mouse may make as many runs as possible.
3. When the Micro Mouse reaches the maze center it may be manually lifted out and
restarted or it may make its own way back to the start square. Manually lifting it out
shall be considered touching the Micro Mouse and will cause it to loose the 10second bonus on all further runs.
4. The time for each run shall be measured from the moment the Micro Mouse leaves
the start square until it enters the finish square. The total time on the maze shall be
measured from the time the Micro Mouse is first activated. The mouse does not have
to move when it is first activated but it must be positioned in the start square ready to
run.
5. The time taken to negotiate the maze shall be measured either manually by the contest
officials or by infrared sensors set at the start and destination. If infrared sensors are
used, the start sensor shall be positioned at the boundary between the start square and
the next unit square. The destination sensor shall be placed at the entrance to the
destination square. The infrared beam of each sensor shall be horizontal and
positioned approximately 1 cm above the floor.
6. The starting procedure of the Micro Mouse shall not offer a choice of strategies to the
handler.
7. Once the maze configuration for the contest is disclosed, the operator shall not feed
the Micro Mouse with any maze information.
8. The illumination, temperature, and humidity of the room in which the maze is located
shall be those of an ambient environment. Requests to adjust the illumination may be
accepted at the discretion of the contest officials.
9. If a Micro Mouse appears to be malfunctioning, the handlers may ask the judges for
permission to abandon the run and restart the Micro Mouse at the beginning. A Micro
Mouse shall not be re-started merely because it has taken a wrong turn.
10. If a Micro Mouse team elects to stop because of technical problems, the judges may,
at their discretion, permit the team to run again later in the contest with a 3-minute
maze time penalty. For example, assume a Micro Mouse is stopped after 4 minutes; it
must be restarted as if it had already run for 7 minutes, and will have only 8 more
minutes to run.
11. If any part of a Micro Mouse is replaced during its performance, such as batteries or
EPROMS, or if any significant adjustment is made, the memory of the maze within
the Micro Mouse shall be erased before restarting. Slight adjustments, such as to the
sensors may be allowed at the discretion of the judges, but operation of speed or
strategy controls is expressly forbidden without a memory erasure.
12. No part of the Micro Mouse (with the possible excretion of batteries) shall be
transferred to another Micro Mouse. For example if one chassis is used with two
alternative controllers, then they are the same Micro Mouse and must perform within
a single 15-minute allocation. The memory must be cleared with the change of a
controller.
13. The contest officials shall reserve the right to stop a run, or disqualify a Micro Mouse,
if they believe its continued operation is endangering the condition of the maze.
Appendix B: Software Flowchart
First Pass
Has square been
visited
Done
no
Is this
the end
Add to
graph
yes
Right
door?
yes
Visited
no
yes
no
Go
right
yes
Forward
door?
Visited
no
yes
Left door?
no
Go straight
yes
Visited
no
yes
Return to
last vertex
and delete
graph
no
Go left
Appendix C: Prototype Motor Driver Assembly Code
ORG 00H
JMP START
ORG 33H
JMP START
START:
RIGHT_WHEEL:
RIGHT_FWD:
LEFT_WHEEL:
LEFT_FWD:
TURN_WHEELS:
LOOP:
END
SETB
MOV
MOV
MOV
ANL
JZ
ORL
JMP
ORL
MOV
ANL
JZ
ORL
JMP
ORL
MOV
MOV
MOV
MOV
JMP
IE.7
;ENABLE INTERRUPTS
R0, #00H
;CLEAR R0
B, P3;
;CHECK DIP SWITCHES
A,B
A, #00000001B
;CHECK S0,RIGHT FORWARD
RIGHT_FWD
R0, #00010000B
;RIGHT WHEEL REVERSE (CW)
LEFT_WHEEL
R0, #00100000B
;RIGHT WHEEL FWD (CCW)
A, B
A, #00000010B
;CHECK S1, LEFT FORWARD
LEFT_FWD
R0, #10000000B
;LEFT WHEEL REVERSE (CCW)
TURN_WHEELS
R0, #01000000B
;LEFT WHEEL FORWARD (CW)
PWCM, #01111111B ;SET PWM COMPARE TO 127
PWMP, #01111111B ;SET PWM PRESCALAR TO 127
P1, R0
;OUTPUT WHEEL DIRECTIONS
PWENA, #00000001B;BEGIN PULSING MOTORS
LOOP
;LOOP UNTIL PWN DONE
;THEN CHECK MOTORS
Appendix D: Micromouse Software
See program code
Appendix E: Dijkstra’s Shortest Path Algorithm
See program code
Bibliography
1) Flynn, Anita M, and Joseph L. Jones. Mobile Robots: Inspiration to Implementation.
Wellesley, Mass: A K Peters, 1993.
2) McComb, Gordon. Robot Builders Bonanza: 99 Inexpensive Robotics Projects. San
Francisco, CA: TAB Books, 1987.
3) Tanenbaum, Andrew. Computer Networks. Upper Saddle River, NJ: Prentice Hall
PTR, 1996