Would this system work?

.:John:.

Would a system like the following work?:

EDIT: FORGET GENETIC ALGORITHMS!!!

Important note: The robot will be sending sensor data to the computer continously, the computer will remember the sensor data throughout the entire command.

The Boe-bot (somehow, I have not decided yet) receives a command to go to the kitchen. The bot then sends a request to the computer for the best route to the kitchen based on the robots sensor input probably using GA(genetic Algorithms). (I think that is possible) The robot then follows that path until the sensors detect an obstacle in the way. The bot sends a request back to the computer for a new instruction set. Computer sends new instructions. This continues until the robot has reached the kitchen.

Is this completely impossible?

P.S. I'm taking this project in steps. GA is first on my list. Although many people have been saying it would be very difficult.

Post Edited (curious_roboticist) : 10/14/2007 8:33:58 PM GMT

Graham Stabler

I don't think you need GAs to find the route, more likely you would just work out the shortest route by searching the route possibilities. I'm sure there are efficient ways to do this, perhaps search for path planning or similar. If it's not just about distance then you can perhaps use some fuzzy logic which is not as bad as it sounds really, it just means making decisions in situations that are not black and white.

Graham

.:John:.

I also have been wondering. I have been looking at using a PING sensor to map the location. One that really seemed like a very good idea was one in the Sandbox. It should be on the first page right now.

Dgswaner

Curious the problem you have is how does you bot know where it is. so when it calls home for the best path, This is the major issue with any navigation, I know it's long but if you read all of Mapping post that you posted on it will explain our thought process and how we got to were we are now. this is also why we chose the Propeller, we feel that it can handle the whole application and not have to be linked to a PC. on a side note... I'm linking my bot to a PC via, the RF modules and a PINK so I can track sensor and position data. but it won't calculate position or anything.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
A complex design is the sign of an inferior designer. - Jamie Hyneman, Myth Buster

.:John:.

Dgs- Ok, I did end up reading the entire post. Most of the people replying said to use external transmitters and RFID chips. I hoped to avoid a system like this. I want the boe-bot (I do not know spin, so prop is out i think) to use one to two ping sensors, a compass , and maybe the color sensor or CMU cam. Plus an RF link to a computer. The bot will continously transmit compass, ping, and video or color data. Then the computer will try to match the current location using a database with old information. The computer will number how many points are the same between current data and old data. If the robot has been in the location before, it will know. If not, the data will be recorded. Distances between points from an old map will then be used to graph the bots location. Hopefully this system will be precise to a few inches.

-CR

parts-man73

Curious - you asked for my help in a PM, twice.... Probably be better to post, and wait for users that have knowledge in the peticular subject that you need help with to respond. I am flattered however that you requested my assistance.

Let's break this down some.

Link to a computer, that's no problem, there are a number of products able to send data wirelessly from a Boe-bot back to a computer.

Computer sending instructions to the Boe-bot to control it's motions. Again, this should be obtainable.

Computer processing the data to create the instuctions, possible, what kind of experience do you have in programming? what language do you plan to use, etc?

Also what is the number of / type of sensors you plan to use? Does the Stamp you plan to use have enough IO pins available to interface the various sensors you plan to use? The processing power to read all sensors, transmit and receive the data, and move the servos?

I would almost think the Propeller would be the better choice, it's multi-cog architecture lends it self well to this type of application.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Brian

uController.com - home of SpinStudio

Mike Green

curious_roboticist,
I think it is possible to do what you want with a Stamp. I don't think it's doable with a plain BS2. You'll need the speed and extra storage of a BS2p-series device. You may need an external serial buffer like the one from Protean Logic depending on how tolerant the PC end is of dropped messages from the PC to the Stamp.

That said, I still recommend using a Propeller. You have a demanding project in mind and the extra processing power and multiple processors will actually simplify your design and programming.

L. Scott Hudson

Hey Curious someone thinks we are the same person...

http://forums.parallax.com/forums/default.aspx?f=10&m=223478&p=1

Maybe we are doing the same thing. Email me offline if you want... we could collaborate. Where are you anyway? I'm in Austin TX.

Scott

.:John:.

Scott-We should collaborate.

Mike- I do not understand the purpose of the prop in this application, pretty much all the BS2 will do is transmit and receive information from the computer.

Parts Man- Sorry about the multiple PMs. I plan on using the compass, ping, and either the color sensor or CMU cam. I should have enough pins for sure.

Post Edited (curious_roboticist) : 10/13/2007 7:54:19 PM GMT

.:John:.

Ok, I will be ordering the parts soon, but before I do, does anyone have any suggestions on what sensors I could use.

Mike Green

The sensors you could use depend on what you want and how that matches the capabilities of the sensors.

Obviously, you could use an ultrasound distance sensor like the PING or an IR distance sensor. Each has limitations in terms of the kinds of surfaces they work well with and the distances they handle. You can use a passive infrared sensor like those used in burglar alarms to detect people.

The CMU-cam allows for detection at a distance, but the analysis of images is complex and the bandwidth to transfer the information is relatively high even with the limited information from the CMU-cam.

.:John:.

Would an onboard computer be a better solution than wireless? I want to use my boe-bot though and a computer might be too heavy.

Post Edited (curious_roboticist) : 10/13/2007 9:06:19 PM GMT

Mike Green

I think you need some practical experience with the hardware before you get too lost in dreaming about what might be possible. Put together a BoeBot, install either a BOE or Propeller Protoboard on the chassis and see what it's capable of. See how quickly it moves, how well you can control its direction and extent of movement, how much weight it can handle, and where the pieces need to be mounted to avoid instability. Experiment with different kind of sensors. I'd start with a PING and an IR detector/emitter. There's existing code for both the Stamp and the Propeller that you can start with. See for yourself how they respond to different surfaces and different distances. If you like the idea of a CMU-cam, get one and see how it handles different shapes. If you don't want to buy too much without having a better idea of what it can do, read articles. Nuts and Volts, Servo, and others are good resources. If you haven't bought the Stamp itself yet and you're sure you don't want to deal with a Propeller, get at least a BS2p or BS2px so you have some extra memory and speed and features.

.:John:.

I have the BOE-bot already, but I probably will buy the BS2p or BS2px. Ping sensor is also on my list

Post Edited (curious_roboticist) : 10/14/2007 7:08:28 PM GMT

.:John:.

I'm not going to use a computer to operate this robot anymore. All the wireless communication systems I saw did not do what I needed them too.

Post Edited (curious_roboticist) : 10/14/2007 7:57:20 PM GMT

Dgswaner

Curious, I have a bot with pretty much every thing your talking about, including an RF link. I can personally assure you that a BS2 will not handle what you are doing, I'm not saying your project isn't doable, but I will say that I agree with mike, I think you should buy the sensors try them out individually, and get familiar with them and then decide how to tackle your project. I went through the same process the possibilities look great on paper but there are so many issues you don't see until you get deep into a project. I decided to go with a propeller rather than try chaining several stamps together to handle the memory space and multi processing needed. I don't know spin and I was able to get my bot servos and motor control working on a propeller in just a couple of days.

As far as on board computer.... I wouldn't even consider it on a Boe bot, I have a Micro ATX computer on board (not a boe bot), about 4"x6" and the battery requirements alone would pop the tires off a boe bot.

I'm not trying to discourage you, quite the opposite I would rather you learn from my mistakes and allow you to get to the finish line without, making the same mistakes I did. when you get working on this project I'd be more than happy to tell you exactly what I did in each step. and you can do with it what you will.

Aim for the stars!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
A complex design is the sign of an inferior designer. - Jamie Hyneman, Myth Buster