Introduction: Maze Solving Robot
Additional information at : http://patrickmccabemakes.com/hardware/Maze_Solvin...
In this instructable I will be showing you how to build a maze solving robot. This is actually my 3rd attempt at making one. The first was a complete failure. The second was alright at finding the end of the maze, but it could not go back and drive the quickest path. This is my current and 3rd one. It is able to find the end of the maze and then, when put back at the start, drive the shortest path to the end of the maze without going down any dead ends. What surprises me is that it really does not mess up.
This was just a short introduction, everything else from theory, building, and programming will be explained in later steps.
Parts list:
- 2 Micro Gear Motors
- Pair of Motor Brackets
- Pair of Wheels
- Ball Caster
- Analog Reflectance Sensor Array
- #2 Bolts and #2 Nuts
- Arduino RBBB
- Motor Driver IC
- 4AAA Battery Holder
- 4 Rechargeable AAA Batteries
General Parts:
- 3 1in Spacers (Home Depot)
- Bolts and nuts that fit the spacers (Home Depot)
- Hook up wire ( I use 22 gauge)
- Solder
- .100" Female and Male headers
- Velcro
Tools:
- Soldering Iron
- Philips Screw driver
- Wire Strippers
- Something to cut the wire with
Step 1: The Theory Part 1
What are the steps In maze solving?
There are basically 2 steps. The first is to drive through the maze and find the end of it. The second is to optimize that path so your robot can travel back through the maze, but do it perfectly with out going down any dead ends.
How does the robot find the end of the maze?
I use a technique called the left hand on the wall. Imagine you are in a maze and you keep your left hand on a the edge of the wall at all times. Doing this would eventually get you out of a non-looping maze. This instructable will only deal with mazes that do not loop back on themselves.
This left hand on wall algorithm can be simplified into these simple conditions:
- If you can turn left then go ahead and turn left,
- else if you can continue driving straight then drive straight,
- else if you can turn right then turn right.
- If you are at a dead end then turn around.
The robot has to make these decisions when at an intersection. An intersection is any point on the maze where you have the opportunity to turn. If the robot comes across an opportunity to turn and does not turn then this is consider going straight. Each move taken at an intersection or when turning around has to be stored.
L = left turn
R= right turn
S= going straight past a turn
B= turning around
So let us apply this method to a simple maze and see if you can follow it. View the photos to see this method in action.
The red circle will be the robot.
As you can see in the photos for this example, the final path is LBLLBSR.
There are basically 2 steps. The first is to drive through the maze and find the end of it. The second is to optimize that path so your robot can travel back through the maze, but do it perfectly with out going down any dead ends.
How does the robot find the end of the maze?
I use a technique called the left hand on the wall. Imagine you are in a maze and you keep your left hand on a the edge of the wall at all times. Doing this would eventually get you out of a non-looping maze. This instructable will only deal with mazes that do not loop back on themselves.
This left hand on wall algorithm can be simplified into these simple conditions:
- If you can turn left then go ahead and turn left,
- else if you can continue driving straight then drive straight,
- else if you can turn right then turn right.
- If you are at a dead end then turn around.
The robot has to make these decisions when at an intersection. An intersection is any point on the maze where you have the opportunity to turn. If the robot comes across an opportunity to turn and does not turn then this is consider going straight. Each move taken at an intersection or when turning around has to be stored.
L = left turn
R= right turn
S= going straight past a turn
B= turning around
So let us apply this method to a simple maze and see if you can follow it. View the photos to see this method in action.
The red circle will be the robot.
As you can see in the photos for this example, the final path is LBLLBSR.
Step 2: The Theory Part 2
Ok so now you have a path. In this case it is "LBLLBSR", but how does the robot change that into the correct path? Well lets take a look at what the correct path would be. Look at the photos for the correct path.
Final correct path = SRR
So we need our path to go from LBLLBSR to the right path that is SRR. To start off we look at where we went wrong. A "B" indicates the robot turned around meaning it went down the wrong path. To optimize the path we have to get rid of the "B" by using some substitution.
Final correct path = SRR
So we need our path to go from LBLLBSR to the right path that is SRR. To start off we look at where we went wrong. A "B" indicates the robot turned around meaning it went down the wrong path. To optimize the path we have to get rid of the "B" by using some substitution.
Step 3: The Theory Part 3
Lets look at the first 3 moves in the path "LBLLBSR". These moves are "LBL".
That move looks like the photo.
Instead of turning left then turning around and turning left again, the robot should have gone straight. So we can say that LBL = S.
This substitution is what the robot uses to optimize the path. That is one example but here is the whole list:
LBR = B
LBS = R
RBL = B
SBL = R
SBS = B
LBL = S
You may not come across all of these when maze solving, but they are required when optimizing the path. Some even put "B" back into the path. This is required to further optimize the path correctly. You can figure out why for yourself or just trust me.
Lets optimize our path now that we know how to:
Path = LBLLBSR
LBL = S so our new path would be: SLBSR
We also know LBS = R so our new path would be: SRR
As you can see we got the path that we were looking for.
My robot optimizes the path as it travels. The path is stored in an array and every time it goes to store a new move, it checks to see if the previous move was a "B", if it was then it optimizes the path. You need to know at least 3 moves to optimize the path: The move before and after the turn around (and the turn around itself).
That move looks like the photo.
Instead of turning left then turning around and turning left again, the robot should have gone straight. So we can say that LBL = S.
This substitution is what the robot uses to optimize the path. That is one example but here is the whole list:
LBR = B
LBS = R
RBL = B
SBL = R
SBS = B
LBL = S
You may not come across all of these when maze solving, but they are required when optimizing the path. Some even put "B" back into the path. This is required to further optimize the path correctly. You can figure out why for yourself or just trust me.
Lets optimize our path now that we know how to:
Path = LBLLBSR
LBL = S so our new path would be: SLBSR
We also know LBS = R so our new path would be: SRR
As you can see we got the path that we were looking for.
My robot optimizes the path as it travels. The path is stored in an array and every time it goes to store a new move, it checks to see if the previous move was a "B", if it was then it optimizes the path. You need to know at least 3 moves to optimize the path: The move before and after the turn around (and the turn around itself).
Step 4: The Theory Part 4
Here is another example.
Using the left hand on the wall algorithm, here is the path the robot would take:
LLLBLLLRBLLBSRSRS
Now here is the process of shortening that path:
LL(LBL = S)LL(RBL = B)(LBS = R)RSRS
The new path would be:
LLSLLBRRSRS
Continue shortening it until all the “B”s are gone:
LLSL(LBR = B)RSRS
The new path would be:
LLSLBRSRS
Continue shortening it:
LLS(LBR = B)SRS
The new path would be:
LLSBSRS
Continue shortening it:
LL(SBS = B)RS
The new path would be:
LLBRS
Continue shortening it:
L(LBR = B)S
The new path would be:
LBS
The final path is:
LBS = R
Using the left hand on the wall algorithm, here is the path the robot would take:
LLLBLLLRBLLBSRSRS
Now here is the process of shortening that path:
LL(LBL = S)LL(RBL = B)(LBS = R)RSRS
The new path would be:
LLSLLBRRSRS
Continue shortening it until all the “B”s are gone:
LLSL(LBR = B)RSRS
The new path would be:
LLSLBRSRS
Continue shortening it:
LLS(LBR = B)SRS
The new path would be:
LLSBSRS
Continue shortening it:
LL(SBS = B)RS
The new path would be:
LLBRS
Continue shortening it:
L(LBR = B)S
The new path would be:
LBS
The final path is:
LBS = R
Step 5: The Design
I first want to say that your robot does not have to built in the same manner mine is. The biggest thing I see is that I designed and laser cut a chassis for my robot. I will post the AutoCAD file, but this is about all I can do. So if you are not able to laser cut the design, do not worry. As long as you have some Arduino, the same sensors, and motors then you should be able to get my code to work on your robot with minor changes. Recently I was just helping some one online who made a copy of mine without laser cut parts. The code ran on his robot with little tweaking.
You can see in the photos the design I made and the parts I used a laser cutter to make. The parts are made from acrylic that is .08" thick. This can be found at The Home Depot.
In the third photo you can see the bottom deck in my 2 deck chassis design. The bottom deck has mounting holes in the rear to bolt the motors on, holes in the front to bolt the ball caster on, a hole to bolt the Arduino on, and 2 tabs in the front where the sensor bolts onto. The bottom deck also has 3 mounting holes that allow it to be connected to the top deck using bolts and spacers.
The top deck has a single, large hole to run wires from the battery pack that is velcroed on top to the electronics on the bottom deck. It also has 3 holes that allow it to be connected to the bottom deck.
It is a simple, compact design that allows for a simple assembly with minimal wiring.
You will see in the next steps how and where the parts fit on.
You can see in the photos the design I made and the parts I used a laser cutter to make. The parts are made from acrylic that is .08" thick. This can be found at The Home Depot.
In the third photo you can see the bottom deck in my 2 deck chassis design. The bottom deck has mounting holes in the rear to bolt the motors on, holes in the front to bolt the ball caster on, a hole to bolt the Arduino on, and 2 tabs in the front where the sensor bolts onto. The bottom deck also has 3 mounting holes that allow it to be connected to the top deck using bolts and spacers.
The top deck has a single, large hole to run wires from the battery pack that is velcroed on top to the electronics on the bottom deck. It also has 3 holes that allow it to be connected to the bottom deck.
It is a simple, compact design that allows for a simple assembly with minimal wiring.
You will see in the next steps how and where the parts fit on.
Attachments
Step 6: Attaching the Motors
In this step I simply attached my motors to the top side of the bottom deck using 2 #2 bolts for each one. You can see that I have a little bit of wire attached to each motor, this is because the motors were removed from my previous maze robot. Then attach the wheels by simply pressing them onto the motor shaft. You have to line the "D" shaped shaft up with hole in the center of the wheel.
Step 7: The Arduino
The first thing you should do in this step is follow the instructions to assemble your RBBB. Those instructions are available here. You will need to snip part of the board off like it says you can do at the bottom of the instructions because you will not be needing it. "If you dont’ need the power jack or regulator just snip them both off. The easiest way to cut the board is with a large pair of tin snips, but hack saws, jewelers saws and band saws also work well." Do not solder any headers onto the board besides the ones used for the FTDI programming port.
Next solder a 9 pin strip of female headers on the left side of the board from pins labeled "5v" to "A0". This will mate with the sensor plug later. Solder a 4 pin strip of female headers on the right side of the board from pins labeled "D5" to "D8". These pins will be used to control the motor controller. Finally solder a 2 pin strip of female headers on the front right of the board on the GND and 5V pins. These will supply power to the motor controller.
You can ignore the fact that the motor controller and top deck in the photos. We will get to those.
Next solder a 9 pin strip of female headers on the left side of the board from pins labeled "5v" to "A0". This will mate with the sensor plug later. Solder a 4 pin strip of female headers on the right side of the board from pins labeled "D5" to "D8". These pins will be used to control the motor controller. Finally solder a 2 pin strip of female headers on the front right of the board on the GND and 5V pins. These will supply power to the motor controller.
You can ignore the fact that the motor controller and top deck in the photos. We will get to those.
Step 8: The Motor Controller
I designed a PCB for my motor controller. I have uploaded the Eagle files for you. I used the Dorkbot service to have them made.
If you do not use a PCB, you can always make this simple circuit on some prototype board.
Now to wire the motor controller up. I have a labeled diagram in the photos. Looking at the last photo, I have the motor on the bottom of the picture connected to the what I have labeled as "M1-A" and "M1-B". These are the outputs of the motor controller for the first motor. I have the motor in the top of the last photo connected to what I have labeled as "M2-A" and "M2-B". These are the outputs of the motor controller for the second motor.
Now time for the inputs of the motor controller. The Arduino digital output 7 gets wired to what I have labeled as "In 1A". This is the first input of the first motor. The Arduino digital output 6 gets wired to what I have labeled as "In 1B". This is the second input of the first motor. The Arduino digital output 5 gets wired to what I have labeled as "In 2A". This is the first input of the second motor. The Arduino digital output 8 gets wired to what I have labeled as "In 2B". This is the second input of the second motor.
Finally Power and ground get connected to the Arduino's 5v and Gnd pins at the front of the board. In the picture I have the wires attached but not plugged in yet.
As you can see I also went ahead and used some of the #2 nuts and bolts to bolt down the ball caster and Arduino.
If you do not use a PCB, you can always make this simple circuit on some prototype board.
Now to wire the motor controller up. I have a labeled diagram in the photos. Looking at the last photo, I have the motor on the bottom of the picture connected to the what I have labeled as "M1-A" and "M1-B". These are the outputs of the motor controller for the first motor. I have the motor in the top of the last photo connected to what I have labeled as "M2-A" and "M2-B". These are the outputs of the motor controller for the second motor.
Now time for the inputs of the motor controller. The Arduino digital output 7 gets wired to what I have labeled as "In 1A". This is the first input of the first motor. The Arduino digital output 6 gets wired to what I have labeled as "In 1B". This is the second input of the first motor. The Arduino digital output 5 gets wired to what I have labeled as "In 2A". This is the first input of the second motor. The Arduino digital output 8 gets wired to what I have labeled as "In 2B". This is the second input of the second motor.
Finally Power and ground get connected to the Arduino's 5v and Gnd pins at the front of the board. In the picture I have the wires attached but not plugged in yet.
As you can see I also went ahead and used some of the #2 nuts and bolts to bolt down the ball caster and Arduino.
Step 9: The Sensor
The sensor comes as an array of 8 sensors. There are 2 on the end that can be removed and need to be removed. I used a pair of side cutters to do this. A 9 pin strip of female headers has to be soldered on to the sensor from "GND" to pin "6" of the sensor. I then plugged in a strip of male headers.
These sensors put out an analog voltage based on how much IR light is being reflected back. We can use these to detect the white and black areas on the maze. The sensor outputs a low voltage near 0V if the sensor detects a white surface. The sensor outputs a voltage near VIN if it detects a dark surface.
These sensors put out an analog voltage based on how much IR light is being reflected back. We can use these to detect the white and black areas on the maze. The sensor outputs a low voltage near 0V if the sensor detects a white surface. The sensor outputs a voltage near VIN if it detects a dark surface.
Step 10: Attach the Top Deck
Bolt the top deck on using the 1in spacers and some bolts and nuts. Use velcro to attached the battery pack to the top deck. Run the leads from the battery pack through the hole in the top deck to the bottom deck. This battery pack already has a built in power switch.
I found it easiest to leave the screw out of the battery pack. The screw prevents the battery pack from opening, but it has clips that do a good job of keeping it closed without the screw. This allows the batteries to easily be changed by sliding the battery pack open. That means you do not necessarily have to lift the battery pack off the velcro to change the batteries.
I found it easiest to leave the screw out of the battery pack. The screw prevents the battery pack from opening, but it has clips that do a good job of keeping it closed without the screw. This allows the batteries to easily be changed by sliding the battery pack open. That means you do not necessarily have to lift the battery pack off the velcro to change the batteries.
Step 11: Attach and Wire the Sensor
The sensor array bolts onto the front of the robot using #2 bolts. The pin on the far left of the sensor is GND and gets wired to GND on the Arduino. The second most left pin is Vcc and it gets wired to the 5V pin on the Arduino. The pins labeled 6-1 on the analog sensor get wired to the Arduino analog pins from 5-0. So pin 6 on the sensor gets wired to the Arduino analog pin 5, pin 5 of the sensor gets wired to the Arduino analog pin 4, etc.
Step 12: Attach Power
The last thing to do is wire the battery pack to the Arduino. This is done by soldering the wires to the power pins at the rear of the Arduino RBBB. The battery pack has a built in power switch so that is why we can simply solder the power leads straight to the back of the Arduino RBBB. The robot is now complete.
Step 13: The Program
I had fun writing this program. Basically there are a few functions that take care of the whole maze solving process. The left hand on the wall function holds the logic to read the sensors and navigate the maze following those rules I outlined earlier. The turning functions are programmed in a fashion so that the robot continues to turn until it sees the black line in which case it then drives forward. There is also a function programmed with a little line following capabilities. It is no PID line following, but does allow the robot to stay on course. This function also controls the robot's speed.
The more detailed function is that of shortening the path. A letter is stored at every intersection that indicates the direction the robot took. If the previous letter was a "B", indicating a wrong turn, the function that shortens the path is run to substitute letters in place of those 3 letter sequences outlined earlier.
Finally there is a function that replays the shortest path that the robot calculated when the robot is picked up and then placed back down on the black line. The robot can detect when it is placed back down and then runs the shortest calculated path.
You can download the code below.