Hello! My K-9 Microchip Pic Autonomous Robot

[dj sures]

Hello,

I'm new the forum . I've been building little robots for quite a while . and recently began building something that demonstrates all i've learned . This is my version of the Doctor Who K-9 . It's pretty smart for having only one sensor to deal with it's surroundings .

sorry for the dark video quality . I'd like to make a new video when it's daytime for you .

[youtube=425,350]uJ_hC8fepOI[/youtube]

[dj sures]

Btw, i should show this video . This is the first version i made of it . It had no body, but it demonstrates the initial ability to understand it's enviroment using the single sharp sensor by scanning and remembering X/Y distances ..

[youtube=425,350]3jmYztwaq7o[/youtube]

[Rebelgium]

Nice work! Tnx for sharing those videos with us

It seems as if your K9 looks downward every 3 seconds or so, but that means if you've got bad luck, it might scan downwards just before a ledge, and then drive on forward and falling off the ledge.
Am I right? or did I miss something there

[dj sures]

it appears it should, but it won't

the reason is timing . i timed the speed of the forward movement with the head scanning . this is how it works .

the looking down is actually a stored array at 8 angles . when the system is first initialized (turned on), it does a 8 point scan. He stores the scan results . Then, each time you see his head move down, he's comparing the results . he's lookin for both raised and lowered surfaces . i also use an algorythm to determine false positives . so the speed of his wheels, and the distance of the scan, prevents him from driving off edges . it's all in with the timing

i also keep track of X scan results . and the speed of which he travels (and direction) allows him to compensate for what to expect . so because of that, he's able to detect motion with the distance sensor . it's pretty simple really, the programming assumed that only living things move . so if something moved, it must be a living thing, so interact with it .

there are a few other tricks i do in programmin to keep track of exit paths . ways for him to get away if he notices a drop-off, raised surface, or trapped corner . so his exit path is always taken .

of course, if you pick him up and move him, then it pretty much screws him up for a while until he's refreshed the arrays .

i'll make a new video this weekend when it's daytime . with more examples of his interaction and enviromental understanding .

[Rebelgium]

good thinking
Really interesting, I'm waiting for you next reply with that video and more code explanation.

[dj sures]

good thinking
Really interesting, I'm waiting for you next reply with that video and more code explanation.

what would you like to know about the code? i have a day of sitting in front of my computer at work, so i can answer any questions ..  So i guess i can give a little rundown of the execution now ..

First the structure . The framework i developed in C has been evolving for a few years . And here is how it works...

Code Framework
Functions are split up into seperate physical files that isolate their purpose. Here is a list of the files and their appropiate functions and parameters..

Head_Servos.c
void moveHead(unsigned int x, unsigned int y)
- The functions keeps track of the last X and Y position so it knows how many pulses to send the servo to get to the next position. The X and Y is currently from 0-16 respectively . I just made that simple because the us pulses for the servos i work with are between 650 and 2250 (1600 difference, divided by 100 gives me 16 points of resolution).

Sounds.c
The sound driver i built oscillates frequency also . Which is why the sounds you hear are very slippery to be more robotic.

Scans.c
Scans are just that. They are functions which can be re-used for different mode applications. The scans populate and react to the global X/Y storage of distances that is accumulated by themselves.

// Returns the closest object distance to the right
unsigned int getMaxRight(unsigned int y)

// Returns the closest object distance to the left
unsigned int getMaxLeft(unsigned int y)

// If an object to the right is closer then MIN_DISTANCE then returns right away
boolean isRightClear(unsigned int xMax, unsigned int y)

// Returns the closest object distance to the left
boolean isLeftClear(unsigned int xMax, unsigned int y)

// Returns the closest object distance in the center
unsigned int getMaxCenter(unsigned int y)

// This must be called before we can check the floor
void initFloorArray()

// Determines if the floor has something on it that it can avoid
// For example, stairs or a shoe
boolean getFloorClear()

// Inits the array for the directionOfDifference
void initDirectionOfDifference()

// Returns the direction that something changed
unsigned int getDirectionOfDifference()

Movements.c
Movements are predefined re-usable macros, so to speak. Mostly they are simply goForward, goLeft, etc. But I also included the Atonomous feature as a movement. So you can specify how many cycles to perform atonomously.

// Set the eyes statuseses
void setEyes(int1 lefteye, int1 righteye)

// Generic flash eyes
void flashEyes()

// Center the head servos
void centerHead()

void goForward()
void goLeft()
void goRight()
void goStop()
void goReverse()

// Reverse, avoid objects and do the funny backup beep
void reverseAndBeep() {

// Avoid an object at the right by going left
void goLeftAvoidObjectAtRight()

// Avoid an object at the left by going right
void goRightAvoidObjectAtLeft()

// Initialize the head servos by stretching their limitations. So we can determine the center
void initHeadServos()

// Perform atonomous exploration for the number of cycles specified
// Each cycle is approx 1 second and 2 inches of travel
void goAtonomous(unsigned int cycles)

[dj sures]

modes.c
Modes encapsulate all other features and puts them into modes. That's what gives him actions for his personality.

// Locate the nearest object and nod and wink at it
void nod()

// Locate the closest object, turn towards it and go to it
void goToClosestObject()

// Locate furthest object and go towards it
void goToFurthestObject()

// Re-init X Y enviroment knowledge
void lookAround()

// Atonomous mode. Explore for an indeterminable ammount of cycles
void atonomous()

// Sleep until motion detected
void sleeping()

// look for motion or change and go towards it
void goTowardsChange()

Menu.c
The menu interface is available mostly eveywhere. I do this by writing my own Delay function, called checkMenu and it accepts a ms delay time. If a button is pressed, then it jumps into the menu interface.

// Check if the menu button was pressed and delay while doing so
boolean checkMenuChangeButton(unsigned int16 delay)

// Check if the menu select button was pressed
boolean checkMenuSelectButton()

// Check if either button was pressed
boolean waitForMenuOrChangeButton()

// Menu mode to show the current adc value
void showADC(unsigned int adcPort)

// Menu option to show the current floor array, compare against stored floor array, position and show difference
void showFloorArray()

// Menu option to show the last change value, current change value, position and the difference
void showLastChange()

// display the menu and stop motors
void showMenu()

// Replacement for delay_ms and checks for menu
void checkMenu(unsigned int16 delay_ms)


util.c
This module contains functions that are considered utilities . I currently only have one because everything else seems to be a scan or a movement . But this is a feature i reuse often. It determines if there is a difference between the value two values by using the a margin . I use this when determing if things have changed

unsigned int1 getDiff(unsigned int origVal, unsigned int newVal, unsigned int diffVal)


There are various other modules too . LCD.C, MoveServos.c, and the main project file.


Neko.h

This is where i configure the specifics of the margins, and limitations, and delays, and global arrays, etc ...

#define RAND_MAX              100

// Analog Ports
#DEFINE ADC_PORT_EYE          0

// Digital Output Ports
#DEFINE LED_STATUS            PIN_C4

// Digital Input Ports
#DEFINE BUTTON_MENU_CHANGE    PIN_C3
#DEFINE BUTTON_MENU_SELECT    PIN_C2

// The ranges of the servo
#DEFINE SERVO_X_MIN           1          // LEFT
#DEFINE SERVO_X_CENTER        10         // CENTER
#DEFINE SERVO_X_MAX           17         // RIGHT
#DEFINE SERVO_Y_MIN           2          // UP
#DEFINE SERVO_Y_CENTER        5          // CENTER
#DEFINE SERVO_Y_MAX           13         // DOWN
#DEFINE SERVO_Y_FLOOR_START   9          // Where he starts looking at the floor for a scan
#DEFINE SERVO_X_ATONOMOUS_MIN 3          // The minimum he looks to the left when moving around
#DEFINE SERVO_X_ATONOMOUS_MAX 15         // The maximum he looks to the right when moving around
#DEFINE SERVO_Y_SLEEP_MAX     12         // How far down to hang his head while sleeping

// The closest we let objects
#DEFINE MIN_DISTANCE          50

// The difference that the floor must change to avoid
#DEFINE FLOOR_CHANGE          15

// The difference of change to wake from sleep
#DEFINE WAKE_CHANGE           13

// The difference of change to notice difference when scanning
#DEFINE DIFF_CHANGE           15

// The Brightness of the LCD
#DEFINE LCD_BRIGHTNESS        20

// GLOBALS
unsigned int FloorArray[SERVO_Y_MAX];
unsigned int OrigDist[SERVO_X_MAX];
unsigned int FloorCheckCount;

[dj sures]

Personality

So this is kinda how it works . Firstly I seed the random generator with the ADC of the distance sensor . The seed determines how active he is .

- The lower the number, the more distance there is means he'll be happier and more exploration.
- The higher the number, the less distance he sees, means he'll be slower and less energetic .

The personality variable is global and changes with activity . If lots of objects are moving, and lots to interact with, then he becomes more interactive and more active .
- If objects are not moving, things are static, then he starts slowing down and sleeping more often, or just sits there looking around for activity .

Modes are enterred and exited based on his activity. He determines how long to execute modes, and when to stop. This even includes the sleep mode. With sleep, he usually only awakes on motion but he will exit the mode if he feels like it . A lot of this is random, but the divider is always changing. So it increases and decreases the resolution (or scale) of the decisions. The larger the scale, the less active he will be . The lower the scale, the more active he will be . It's all determined by a divisor (his personality variable)

[Rebelgium]

thanks for all that info, it has given me more inspiration for my current robot project. 
If you have more time and info, plz be my guest and post it 

[dj sures]

thanks for all that info, it has given me more inspiration for my current robot project. 
If you have more time and info, plz be my guest and post it 

no problem

i don't know what more i could say about this project . today i added a waggin tail, and ears . the ears spin by pager motors . and the tail wags via a servo . i painted K-9 on his side . he's happy now

but i have a damaged servo on his neck cause i dropped him

i'll have to buy a new servo tommorow for the video

[dj sures]

Here is the most recent video..

[youtube=425,350]uJ_hC8fepOI[/youtube]

[Ro-Bot-X]

This is so Cool! Thanks for sharing!
Nice job! You deserve the Grand Premium!
A lot of inspiration for my work! I am currently brainstorming a new robot and I have to say that you gave me the ideea how to scan and move around. I'll give credits for it.
I loved the movie, but text is fading too fast to read it, had to pause to read.

I think that all it needs now is another servo for the mouth, so it can play fetch! Maybe a PIR sensor to find you?

Wonderfull!

[dj sures]

This is so Cool! Thanks for sharing!
Nice job! You deserve the Grand Premium!
A lot of inspiration for my work! I am currently brainstorming a new robot and I have to say that you gave me the ideea how to scan and move around. I'll give credits for it.
I loved the movie, but text is fading too fast to read it, had to pause to read.

I think that all it needs now is another servo for the mouth, so it can play fetch! Maybe a PIR sensor to find you?

Wonderfull!

Hey thanx a lot for the kind words . that means a lot! i can't wait to see what you come up with

[Admin]

Cute! I like the ears and tail, nice touch

You should add my stampy algorithm to it so that it chases stuff!

[dj sures]

Cute! I like the ears and tail, nice touch

You should add my stampy algorithm to it so that it chases stuff!

neat . our minds think alike .. i have another robot i made that does the same thing . it searches for edges . i have it search for any edge (closest edge) .

i was going to do something like that with k9 but instead i went a different route . last night i added a IR detector .. i've never used one before, and i was surprised at how easy it was to implement . i'm using 200k resistor . i tried 300k but it was too sensetive and was picking up all sorts of ir . i find the 200k work well , i get about 4 feet of range .

so the idea here isn't too pick up a specific ir . it's to pick up any ir .. so any remote control at all .. when he sees the IR, he goes towards it . kind like a beacon .

on the 877a, i use the Timer1 for the motor servos . and i use the Timer2 for the IR detector . it's set for a 10ms interrupt . every 10 ms it checks for an IR signal . if it sees one, then he'll turn towards it .

what's cool, is that it interrupts any mode/action he's currently doing . so if he's just nodding, or lookin around, then he'll phsycially turn towards the ir .

if he's in autonomous mode, then he'll turn towards it and continue moving at the source .

so in autonomous mode, using the IR works like a leash . he can follow me around the house .

my next step is to make a IR Ball that he can chase . i'll throw a IR transmitter in a clear plastic blow-up toy ball . and he can chase it around . when he gets it, he can push it away with his nose and chase it again

i'll take a video of the chasing IR signal soon .

[Admin]

It might get a bit complicated when you try to combine obstacle avoidance with ball chasing . . . what happens if a chair leg is between the robot and the ball? This is where your mapping stuff will come handy . . .

last night i added a IR detector .. i've never used one before, and i was surprised at how easy it was to implement . i'm using 200k resistor . i tried 300k but it was too sensetive and was picking up all sorts of ir . i find the 200k work well , i get about 4 feet of range .

what parts you using? that sounds too good of a range . . . is the IR emitter just very bright?

to optimize for the best resistor value, use this equation:
optimized_resistor = sqrt(R_dark*R_bright)
where R_dark is the resistance of the IR detector without an emittor, and R_bright, is the resistance of the IR detector close to the emitter. use a multimeter of course . . .

its the same way you optimize the paired resistor for a photoresistor sensor:
http://www.societyofrobots.com/schematics_photoresistor.shtml

[Asellith]

First of amazing robot! very inspirational. I do have a few questions. The PIR does it get interference from the IR distance sensor? Also do they work well outdoors? I want to do make an automated turret that tracks a person in 2D. For my R2-D2 project. The other question involves another star wars droid. Before I get to far into my R2 project I want to get more experience with sensors so I was going to start with a mouse droid that basically does most of the stuff your K-9 does. So I was curious about the programming to get him to follow you with the PIR. I'm assuming you just locate the larges infrared source and lock onto it them move forward and keep checking it to make sure its still there. but like I said I don't have much experience coding for sensors yet so any incite would be great:)

Thanks
Johnny B

[dj sures]

It might get a bit complicated when you try to combine obstacle avoidance with ball chasing . . . what happens if a chair leg is between the robot and the ball? This is where your mapping stuff will come handy . . .

last night i added a IR detector .. i've never used one before, and i was surprised at how easy it was to implement . i'm using 200k resistor . i tried 300k but it was too sensetive and was picking up all sorts of ir . i find the 200k work well , i get about 4 feet of range .

what parts you using? that sounds too good of a range . . . is the IR emitter just very bright?

to optimize for the best resistor value, use this equation:
optimized_resistor = sqrt(R_dark*R_bright)
where R_dark is the resistance of the IR detector without an emittor, and R_bright, is the resistance of the IR detector close to the emitter. use a multimeter of course . . .

its the same way you optimize the paired resistor for a photoresistor sensor:
http://www.societyofrobots.com/schematics_photoresistor.shtml

ball chasing is done . i'll make u a vid .

it was hard . but a few robots ago i created a function called "goAtonomous"

and you can specify the number of cyles it can go . it starts by going forward, after you've positioned the direction he wants to be in . he'll go for the number of cycles while he avoids things .

meanwhile, back in the bat cave.... the timer2 interrupts and checks for an IR signal . when it finds one, it takes the number of points his head is turned, and moves the body respectively to face the IR source . if the IR is ever in front, then it does nothing . it only turns the body towards an IR signal .

so constantly he turns towards any IR source . it doesn't matter what he's doing . wether he's sleeping, or anything . he'll just turn towards it . the only time he'll ever chase it, is if he's in atonomous mode .in which case his new direction faces the IR source .

if the IR source changes, then he'll see it on the left or right and again reposition himself towards it .

he will never do anythin about an IR source that is directly in front of him (ie when his head is positioned SERVO_Y_CENTER +- 1)

so now he drives around like he usually did, but if there is an IR source, he positions himself at it . it's workin great . i'm having a lot of fun with different remote controls for my tv's, stereo's etc making him come towards .

he's far more interactive now . when you're changing channels or volume on the tv, he comes towards you to see what's up

[dj sures]

First of amazing robot! very inspirational. I do have a few questions. The PIR does it get interference from the IR distance sensor? Also do they work well outdoors? I want to do make an automated turret that tracks a person in 2D. For my R2-D2 project. The other question involves another star wars droid. Before I get to far into my R2 project I want to get more experience with sensors so I was going to start with a mouse droid that basically does most of the stuff your K-9 does. So I was curious about the programming to get him to follow you with the PIR. I'm assuming you just locate the larges infrared source and lock onto it them move forward and keep checking it to make sure its still there. but like I said I don't have much experience coding for sensors yet so any incite would be great:)

Thanks
Johnny B

the IR detector and the IR Distance sensor do not conflict . because i dropped the sensetivity on the IR detector and positioned it under the head facing forward . i'll show u pics tommorow .

i don't check to see if the IR source is still there . i only care if i see it on the right or left, then i position myself respectively . otherwise, i continue going straight and avoiding objects .

and thank you very much for the compliments .

[dj sures]

It might get a bit complicated when you try to combine obstacle avoidance with ball chasing . . . what happens if a chair leg is between the robot and the ball? This is where your mapping stuff will come handy . . .

last night i added a IR detector .. i've never used one before, and i was surprised at how easy it was to implement . i'm using 200k resistor . i tried 300k but it was too sensetive and was picking up all sorts of ir . i find the 200k work well , i get about 4 feet of range .

what parts you using? that sounds too good of a range . . . is the IR emitter just very bright?

to optimize for the best resistor value, use this equation:
optimized_resistor = sqrt(R_dark*R_bright)
where R_dark is the resistance of the IR detector without an emittor, and R_bright, is the resistance of the IR detector close to the emitter. use a multimeter of course . . .

its the same way you optimize the paired resistor for a photoresistor sensor:
http://www.societyofrobots.com/schematics_photoresistor.shtml

BTW thanx for that IR assistance .

Here's a quick video i threw together for you ..

[youtube=425,350]7i8ZP9aJMmc[/youtube]

[dj sures]

ah he's been upgraded in the brain . i have lots of lcd debugging which uses a lot of rom . plus the menu interface for diagnostics, and such . so the old 877a program memory was too small

i upgraded today to a 18F4685 . i'm running it still at 20mhz, but it's a pretty powerful chip, using only 9% of rom

[dj sures]

last night i filled the eeprom with 1024 bytes of 1 bit audio . now he can bark

tom baker asked k-9 to bark once . k-9 said he wasn't programmed to bark .

my k-9 is

[Admin]

soooo . . . ummmmm . . . what does a robot dog sound like?

[dj sures]

soooo . . . ummmmm . . . what does a robot dog sound like?

hehe , i'll make u a vid tom

i just did some documentation to him on my website @ http://atomic.speculation.org/robots

[paulstreats]

to satisfy my curiosity....

for your sound are you using:
 a resistor & capacitor network
 a resitor & transistor network
 or some cleverly elaborate d/a converter - voltage to freq converter - op amp -
 or something else

 I'm hoping that the bark is going to sound authentically 1 bit retro robotics style