Autonomous Cars
08-19-2004, 12:04 PM
#1
Autonomous Cars
Hello RCTech.net,
Months ago I came here asking for help designing a car since our spec rules require 1:10th scale RC car chassis. I ended up getting some valuable information relating to high speed response servo systems and tire selection. All of which helped during my robotics competition.
For those of you interested in robotics, or have been following my adventures learning about RC cars in order to make our robots go a little bit faster, this may be a treat for you. The videos just became available, so I'm sharing them with you.
Our team from UC Berkeley was able to take the 2004 Championship in May. Other schools from around the state and Canada competed in this competition.
The cars are fully autonomous, meaning you put them down, flip a switch, and they follow the track. No remote controlling is allowed. The fastest cars are the ones with the best combination of electronics on them. Everything except the car hardware and the CPU mainboard were designed and produced inhouse. High-power amplifiers, filters, speed controllers, and power distributions were all "homemade."
Although the teams from UCLA had suspension (I believe they were running older Tamiyas), Berkeley teams were all run on "pan cars." I had attempted to migrate our electronics onto a TC3 but did not make it in time. Tuning would have taken more than a week for a new car so the TC3 was out of the question when it came down to the wire. It's okay, because we still WON.
Here's my team's car:
http://www.ece.ucdavis.edu/natcar/images/UCB3-04.wmv
This video is of one of the slower cars from Berkeley who finished 5th place. You can see the difference between a good autonomous vehicle and a mediocre one. Precision is the name of the game.
http://www.ece.ucdavis.edu/natcar/images/UCB4-04.wmv
*Our winning autonomous car travelled at 10feet per second or 7mph. Which is slow compared to how fast our cars can go but you've got to compare the accuracy of the robotic car to the accuracy of our RC cars. One technique every team has considered, but failed, is track memorization. This is one disadvantage that our robots have right now in comparison to humans. Whereas a human can remember pretty well how many turns there are and how fast he can go, these robots are figuring it all out "on the fly." Hopefully next year with some better electronics, the cars will have enough computing power to remember a whole track with enough resolution (0.5 centimeters is barely sufficient --- millimeters is preferable).
Enjoy!
Months ago I came here asking for help designing a car since our spec rules require 1:10th scale RC car chassis. I ended up getting some valuable information relating to high speed response servo systems and tire selection. All of which helped during my robotics competition.
For those of you interested in robotics, or have been following my adventures learning about RC cars in order to make our robots go a little bit faster, this may be a treat for you. The videos just became available, so I'm sharing them with you.
Our team from UC Berkeley was able to take the 2004 Championship in May. Other schools from around the state and Canada competed in this competition.
The cars are fully autonomous, meaning you put them down, flip a switch, and they follow the track. No remote controlling is allowed. The fastest cars are the ones with the best combination of electronics on them. Everything except the car hardware and the CPU mainboard were designed and produced inhouse. High-power amplifiers, filters, speed controllers, and power distributions were all "homemade."
Although the teams from UCLA had suspension (I believe they were running older Tamiyas), Berkeley teams were all run on "pan cars." I had attempted to migrate our electronics onto a TC3 but did not make it in time. Tuning would have taken more than a week for a new car so the TC3 was out of the question when it came down to the wire. It's okay, because we still WON.
Here's my team's car:
http://www.ece.ucdavis.edu/natcar/images/UCB3-04.wmv
This video is of one of the slower cars from Berkeley who finished 5th place. You can see the difference between a good autonomous vehicle and a mediocre one. Precision is the name of the game.
http://www.ece.ucdavis.edu/natcar/images/UCB4-04.wmv
*Our winning autonomous car travelled at 10feet per second or 7mph. Which is slow compared to how fast our cars can go but you've got to compare the accuracy of the robotic car to the accuracy of our RC cars. One technique every team has considered, but failed, is track memorization. This is one disadvantage that our robots have right now in comparison to humans. Whereas a human can remember pretty well how many turns there are and how fast he can go, these robots are figuring it all out "on the fly." Hopefully next year with some better electronics, the cars will have enough computing power to remember a whole track with enough resolution (0.5 centimeters is barely sufficient --- millimeters is preferable).
Enjoy!
08-19-2004, 01:57 PM
#2
that's really cool. great job. how did you track the white line? like what kind of sensor did you use?
08-19-2004, 02:52 PM
#3
Congratulations!
Those are some sweet videos! Was that like a senior project for you at school, or kind of an ongoing "club" project? Things like that can make school actually fun. At least I thought the SAE (Society of Automotive Engineers) Mini-Baja project I did for my senior project was fun (though a heck of a lot of work). I'm sure I speak for the rest of the engineers here (and a lot of the other techno geeks) when I say more details on the cars would be really cool. And congrats on your win!
08-19-2004, 02:58 PM
#4
Good job, that is pretty cool!
08-19-2004, 03:07 PM
#5
Simply Awsome!!!!!!
I am leaning towards robotics myself (trying to learn Stamp Basic and trying to get a hold of a couple of Lego Mindstorms 
) I watched you guys go at it with the autonomous helicopters on the Discovery Channel! That is probally why I now own two helis!
I was wondering how hard was it to design programs for r/c vehicles (both cars and helis) to become fully autonomous? What other gear did you use? I'll just ask these two questions for now, for I'll ask more later. 
) I watched you guys go at it with the autonomous helicopters on the Discovery Channel! That is probally why I now own two helis! I was wondering how hard was it to design programs for r/c vehicles (both cars and helis) to become fully autonomous? What other gear did you use? I'll just ask these two questions for now, for I'll ask more later.
08-20-2004, 09:35 AM
#6
Hi guys,
Glad to see there's *some* interest in robotics on this forum
The white line was a reflective sort of tape. Heh actually it was electrical tape. Under the line was a wire running 100mA of current @ AC 75Khz. Not much.
Teams can use any method they want to find the "track" but spec rules determine that at no point in time will your car drift further than 18 inches from the main line of the track. If you can imagine bombing down a straight section and then suddenly coming up on a turn, this is pretty hard. This was the reason behind the cones in the video, and also the placement of the cones were designed to catch cars in the easiest places to make mistakes. Every cone hit added 2 seconds to your time trial.
Our car did the course in 25 seconds and the 2nd place team did the course in approx 26.5 seconds, just to let you know how devastating hitting ONE cone would be to our team.
The Berkeley teams all used radio frequency sensing of the track. We used tiny antennae to pick up the amplitude of the oscillating current in the wire (through magnetic field). Then onto some amplifiers that we built. The challenge with all the electronics is to make it high-power, and FAST. The faster your stuff is, the quicker your robot can use the data and respond. Our CPU was running at 20MHz, which is mildly respectable for a microcontroller but surely not the fastest out there.
This competition is part of a senior project course at Berkeley, but other schools may treat the competition as a club event. Personally, I am a graduate student at the university and did the competition for fun, since I never had a chance to delve into robotics as an undergraduate. I must say it was a life-changing experience and I highly recommend anyone who has interest in robotics to simply JOIN A COMPETITION, and THEN build your robot.
As far as "how hard" the programming was, the microcontrollers usually come with software kits to do such things. I am not the expert in our software setup since I was more of the circuit designer, but I can tell you a few things. The software was a dialect of C. The software we used was called Keil Microvision and DaVE (Siemens Digital Application Virtual Engineer). DaVE is a piece of shit, avoid using it at all costs as it only sets up certain initializations for you. Once my team figured out how to dump DaVE, we never used it again.
If you understand how to write a program and use: printf, updating registers, hex numbers, and know how to read a manual, you will be fine. Start with Lego Mindstorms and then pick up a Motorola microcontroller through Digikey or Jameco. They actually make kits. This year I may try to form another team to do autonomous helicopter competitions. I can honestly say, I love robotics.
Best of Luck and please let me know if there are any other questions. I leave you with some pictures of my TC3 in robotic trim. Too bad I never got to run it.
By the way, the competition is called: NATCAR
Other robotics competitions are:
Autonomous Helicopter
Micromouse Competition
Donovan
Glad to see there's *some* interest in robotics on this forum
The white line was a reflective sort of tape. Heh actually it was electrical tape. Under the line was a wire running 100mA of current @ AC 75Khz. Not much.
Teams can use any method they want to find the "track" but spec rules determine that at no point in time will your car drift further than 18 inches from the main line of the track. If you can imagine bombing down a straight section and then suddenly coming up on a turn, this is pretty hard. This was the reason behind the cones in the video, and also the placement of the cones were designed to catch cars in the easiest places to make mistakes. Every cone hit added 2 seconds to your time trial.
Our car did the course in 25 seconds and the 2nd place team did the course in approx 26.5 seconds, just to let you know how devastating hitting ONE cone would be to our team.
The Berkeley teams all used radio frequency sensing of the track. We used tiny antennae to pick up the amplitude of the oscillating current in the wire (through magnetic field). Then onto some amplifiers that we built. The challenge with all the electronics is to make it high-power, and FAST. The faster your stuff is, the quicker your robot can use the data and respond. Our CPU was running at 20MHz, which is mildly respectable for a microcontroller but surely not the fastest out there.
This competition is part of a senior project course at Berkeley, but other schools may treat the competition as a club event. Personally, I am a graduate student at the university and did the competition for fun, since I never had a chance to delve into robotics as an undergraduate. I must say it was a life-changing experience and I highly recommend anyone who has interest in robotics to simply JOIN A COMPETITION, and THEN build your robot.
As far as "how hard" the programming was, the microcontrollers usually come with software kits to do such things. I am not the expert in our software setup since I was more of the circuit designer, but I can tell you a few things. The software was a dialect of C. The software we used was called Keil Microvision and DaVE (Siemens Digital Application Virtual Engineer). DaVE is a piece of shit, avoid using it at all costs as it only sets up certain initializations for you. Once my team figured out how to dump DaVE, we never used it again.
If you understand how to write a program and use: printf, updating registers, hex numbers, and know how to read a manual, you will be fine. Start with Lego Mindstorms and then pick up a Motorola microcontroller through Digikey or Jameco. They actually make kits. This year I may try to form another team to do autonomous helicopter competitions. I can honestly say, I love robotics.
Best of Luck and please let me know if there are any other questions. I leave you with some pictures of my TC3 in robotic trim. Too bad I never got to run it.
By the way, the competition is called: NATCAR
Other robotics competitions are:
Autonomous Helicopter
Micromouse Competition
Donovan
08-20-2004, 09:36 AM
#7
Yes, the sensor bridge was made out of carbon fiber, and YES we made it ourselves
PS. Some teams used optical detection in the form of photodetecting diodes arranged in an array on the front bumper of the car. One really innovative team used a CAMERA!!!! to see the track. Their idea was awesome, but unfortunately it takes so much computing power that their car was slow and when they tried to speed it up, it got confused very easily.
Enjoy
PS. Some teams used optical detection in the form of photodetecting diodes arranged in an array on the front bumper of the car. One really innovative team used a CAMERA!!!! to see the track. Their idea was awesome, but unfortunately it takes so much computing power that their car was slow and when they tried to speed it up, it got confused very easily.
Enjoy
Last edited by donoman; 08-20-2004 at 09:42 AM.
08-20-2004, 09:38 AM
#8
Other cars at the competition. Our car is the 2nd one from the front with the white bumper. Ever seen so many wires?
08-20-2004, 10:25 AM
#9
Tech Addict
Man...thats awesome.....now if i can put that tape on the road between Hooters Bar and my house I can drink, ride my car & get home without driving drunk !!!!!! Oh yes, can you please send me the diagram on how to make that robot...so i can stick it in the front of my toyota?....just kiddin
That ROCKS.....the video is just phenomenal.....and the mind behind the project is just plain......well.......Einstein to say the least. That is the closest i can think of to "Back to the Future" thingy. LOL
CONGRATULATIONS to you & your team. Keep up the good work.
BEAM ME UP , Scoty.....
Henry E.
That ROCKS.....the video is just phenomenal.....and the mind behind the project is just plain......well.......Einstein to say the least. That is the closest i can think of to "Back to the Future" thingy. LOL
CONGRATULATIONS to you & your team. Keep up the good work.
BEAM ME UP , Scoty.....
Henry E.
08-20-2004, 11:36 AM
#10
Tech Master
Kondo Robots
Hey guys, I know that it is a little off of the subject but it's robotics.
KO PROPO Japan has formed a seperate division called Kondo. This division specializes in bi-pedal walking robots. This is only available in Japan at the moment. They compete in the Robo-One tour in Asia and Australia.
Blue-tooth controlled, programming required (C+). website video links
KO PROPO Japan has formed a seperate division called Kondo. This division specializes in bi-pedal walking robots. This is only available in Japan at the moment. They compete in the Robo-One tour in Asia and Australia.
Blue-tooth controlled, programming required (C+). website video links
08-20-2004, 01:42 PM
#11
Wow those are very cool.
08-20-2004, 06:03 PM
#12
Re: Kondo Robots
Originally posted by KO PROPO Staff
Hey guys, I know that it is a little off of the subject but it's robotics.
Hey guys, I know that it is a little off of the subject but it's robotics.
J/K Happy Friday Row-bear-toe
08-23-2004, 10:20 AM
#14
Tech Master
Re: Re: Kondo Robots
Originally posted by rtypec
You guys can make all singing all dancing killer robots but you can't make my speedo play "Funky-Town" when I turn it on?
J/K Happy Friday Row-bear-toe
You guys can make all singing all dancing killer robots but you can't make my speedo play "Funky-Town" when I turn it on?
J/K Happy Friday Row-bear-toe
08-23-2004, 08:47 PM
#15
Tech Adept
Re: Autonomous Cars
wooowww !!!! So these actauly follow the track without remote control. Thats crazy !!!