Great Start, Omer! Also looking forward to details. Used Samoan to bypass the Petoi Spam Filter! RZ uses a math hack for the purpose, that does not work for prime numbers of words. Amataga Lelei! Vaavaai atu i faamatalaga auiliili.

Hi Briquila, You may want an MUX (multiplexor) module to extend your I/O channels. From there its simple to access many sensors in a fast loop, but Mind+ alone is not enough, start to cut and paste code any leading GPT gives you, and when it does not work the first time go back for rewrites until it does. It's a fast way to learn anything you want with Arduino or Pi This video covers an Arduino MUX- Arduino Multiplexor - MUX INPUTS & OUTPUTS - ANALOG and PWM (youtube.com)

Hi Reid, True, with such high frame rates video resolution must trade away. 8f/s would also help conserve bandwidth. Fast low-res bursts and high-res frame grabs can happen in sequence, sensitive to context state. Am also setting up vision with Pi as host and Pico Pi camera client streaming the high-def "data firehose" to an external drive, for later rich data training, and only sample select frames in real time, for Pi to serve to the cloud for AI vision MoE (mixture of experts), to return high level contextual data and make apt conversation. Using Landing AI as Machine Vision provider, everyone at breathless pace to constantly upgrade. Stereo "DataEyes" prototypes coming along, that for Bittle need to become 1/10 scale and cheap, where things are "headed" (pun). Saw a new Petoi injection molded cat model with glowing eyes, that could serve in toy dataeye mode, for machine vison with primitive embedded optical communication, as great STEM now, soon to be THz performance. Will post new Vision topics soon, with ample details!

This Forum is at the AI Frontier all robotics is at. Learn all you can about semantic AI, where Natural Language Prompting is all-powerful. Learn to use Large Context Windows to turn any of the major AIs into custom Agents. Create self-sustaining recurrent Prompting Loops of sensor, reasoning, and actuation data. The Session snippet given is both trivial and profound, and reflects the state-of-the-Art. Besides Prompting, the only thing a developer will have to do is interconnects. No more software hassles to patch after the code-free revolution happening now. Existing code-bases and human initiative is a brief advantage. Soon Bittle will get better by itself and be smarter than any of us. Bittle will be helping us navigate life. Swarms of Bittles will acting like big ants able to do most things we do, if the head swivel axis is reworked as a gripper. A workspace can be covered by Machine Vision feedback cameras, no need for eyes on every puppy.

Have been considering buying a box of Bittles, having before bought servos by-the-box. Considerable volume still left before Bittles pack like rectilinear solids. The coolest demo will be Bittles packing and unpacking themselves. Perhaps a next design iteration is to have passive bump-and-cup alignment-points, or even some "lego-lock" capability that allows the bots to comprise meta-structures. Petoi can develop endless add-on features as profit drivers, like a Barbie sale gets lots of knock-on sales of whatnot. Technical users may not favor the dog-head look, but prefer some sort of bus-bar rail to mount sensors along, plus the protective roll cage, etc. Enough of a user base will allow for a third-party accessory market to develop. Furry suits will be popular.

To recognize a color, the lighting must be consistent, and an ROI (region of interest) often mapped. A live image may be matched against a reference image or a look up table, as RGB commonly. The math function can be relaxed to cover related colors in color-space. Alternatively, one might train or simply prompt a suitable GPT. Colorfilter/photo-detector and spectroscopy principles once served as well, and still find uses.

Progress coming fast, from every direction. Programmable and fixed signs will be for robots and people, for situated messaging like teleprompting or advertizing works. Robots will soon easily read screens or blackboards for prompting. Certainly, autonomous driving cars must read and obey signs. For Bittle, the capability can start as simple as sending images to an OCR (optical text recognition) server that then sends semantic output to an LLM generating behavior scripts made of Bittle control primitives, responding intelligently to every kind of semanified input data. Tried two online OCR servers, and the second returned a correct semanification, but a realistic sign in odd light at skewed angles is harder- add your own speed

Cell Phones offer tremendous COTS TRL9 capabilities for robots, like GPS Navigation and QR Codes. It is essential to structure the robot's environment just as traffic signs and signals work or how one finds a flight at an airport. A current bottleneck is Bittle's ancestor, Spot, has trouble with doors. They are so varied by lighting, color, size, and hardware. A QR Code that any cell phone can read could make any door easier to operate. QR codes can lead to a large data files or access AI, so the format is Turing Complete.

USB data cable onboad, wifi and/or cell network for LAN/WAN. Bluetooth good peer-to-peer when robots interact, emitter-detector pairs frontally aimed. Door-cam eyes I am using are wifi.

Further elaboration of concept for a small cell phone to augment Bittle-