Forgot to mention, my end goal is to build multiple battle bots, and have them play a 'knock the opponents out of the arena' game.

Myself and a few friends will take 1 bot each (each identically built) and code our own AI for them to try to win.

>>993712
>Is Raspberry Pi my best option?
It's overkill for the type of project you're describing and will be more difficult than Arduino or other microcontrollers (PIC, ARM, etc). A PI is a full fledged computer where you write software that goes on top of an OS. That gives you some nice features like a file system and multitasking, but now you've got to deal with getting the OS configured to let you use GPIO pins, launch your program on startup, etc. It also draws a lot of power which can be an issue for your battery operated project.

>>993722
Thanks. So you'd recommend Arduino?

Yeah have a look at this kit: https://store.arduino.cc/product/A000078 its one of the best starter kits, search on the net for open code, it is one which has most things

if you are looking for an above starter pro thing, have a look at this: https://www.autonomous.ai/deep-learning-robot this is a very cool machine, high end parts, but also much more expensive.

Last: have a look at opencog.org this is a open source AI framework, being developed by ben goertzel and his team, look him up on youtube I'd say.

Bump (contact) sensors are your friends. I did robotics in high school, we had sonar sensors, IR sensors and cameras, but bump sensors were the only reliable things we could get to work. Especially for a beginner, they will help you actually get some working programs. The sad truth about robotics is that it will never work as well as video game AI because of the uncertainties involved in controlling a physical entity and sensing a physical environment.

That being said, at the time I did not have any knowledge of machine learning. Learn about machine learning. It can be a heavy subject, but it is a fundamental of robotics.

If your interest is more in battling robots than programming AI, I would strongly recommend RC robots. Programming AI without lots of knowledge about the field involves a ridiculous amount of debugging and trial and error. 98% of your time will be spent debugging, for reasons like one of the identical motors on your robot running slightly faster than the other because it has been run for 10 minutes longer.

Anyway, that's just from my experience when I didn't know about machine learning. You may have a better time if you know more than I knew (or know).

Also, Lego technic makes good building materials. Buy your own servos and motors though, and don't be afraid to combine them with glues and other materials.

Tldr bump sensors legos

>>993712
>> AI for videogames
Then build something that can run ROS. Once you have something that can do SLAM, obstacle avoidance becomes path planning.

>>993867
That deep learning robot is way over priced. It's just a turtlebot with a tegra and software installed. If you must buy a robot base get a Turtlebot 2.

Or if you're /diy/ you can build essentially the same thing from an irobot create, a minipc, a kinect, and wood:
http://www.transistor.io/homebuilt-turtlebot.html

>>994643
not op but how accurate are the temp sensors for arduino, is it up to the sensor to get accurate measurement or the sensor itself?

>>993712

If you want some real advice, don't do that

it's THE hack (as in "talentless hack" ) project.

My advice is to think of something original. You should work on a project that is challenging enough to satisfy your dream, but not robotics. Robots are kind of a meme.

>>993722

Don't listen to this. Overkill is fine. A raspberry pi is the best option compared to dealing with tarduinos or embedded AVR shit for any computationally intensive application.

A pi zero is roughly 60 times faster than a 328p clocked at 16 mhz. And that's only if you are smart enough to not use the Arduino bootloader.

>>994780
yeah too bad the Pi cant even handle 1 servo without jittering.

>>994790
The pi is fast enough you can use it as an FM audio transmitter just by putting a wire on GPIO. If you use python you get jitter.

A beaglebone black is superior to the pi in this regard, it has hardware PWM.

>>994790
Your program is sharing the CPU with the OS and other applications. If you implement PWM in code it won`t be stable unless you handle it with timing threads and tell the OS to give your application priority. This is another example of why simple things are better done on a microcontroller, not an application sitting on top of an OS.

>>994777
Not sure exactly what you meant by your question because of typos, but I assume you meant "is it up to the ardiuno or the sensor?"

Also not sure about "temp sensors". If you mean temperature sensors, I've never used them, but I imagine you get what you pay for in terms of accuracy. If you want a sensor with perfect accuracy it will likely cost a lot of money. Just a guess, though.

If you meant bump sensors, they are 100% accurate, as they are a mechanical switch.

As for the arduino/sensor accuracy problem, I don't think that the micro-controller or computer you are using with the sensor should impact your accuracy. It will more likely matter how you collect data from the sensor in your program.
I'm actually talking out of my ass with the "microcontroller doesn't matter" business, though. That's just what makes sense to me. I guess it's possible that different controllers might have more electrical noise when using analog sensors, but this seems a bit silly to me.