T5 Robot construction question on senses

[Baroun Tardis]

Wanting to verify I'm understanding something, because it doesn't feel right

"Creating robot sense" seems to imply that for a robot to be able to see and hear commands as well as a human would, we're looking at:

Vision: 160kcr for a constant-16 that makes it about as observant as a person. Then another 10kcr for the RGB band, total of 170kcr

Hearing: 160kcr for the C-16. 6kcr for 6 freq's. Another 6kcr for voice freqs, total of 172kcr

OK, 172kcr seems somewhat steep for what's basically a Google Home / Alexa bolted on the robot.
Ditto 170kcr seems like a lot for a webcam. People are doing things right now where they run OpenCV on an old cell phone and achieve decent functionality.

Am I misreading how the sensor pricing is supposed to work?
Is there more to the price than what I'm seeing?

 

[Baroun Tardis]

Fiddling with it more -
Is the price the software for handling the incoming sense data?

Vision C=16 is human, and means most everything within 150m is noticed for size bigger than range.
If I'm building a robot for operation inside a ship, for example, I can drop to C=9, and it'll usually notice what's going on within 50m (which is as far as a hallway will likely be) and is pretty much guaranteed to notice everything within 5m (the compartment it's in)... and that trims 70kcr off the robot's price

 

[RogerD]

I haven‘t yet dived into this system, but I will note that these days voice recognition is typically done in the cloud, so there could be larger computers behind it. Not sure what is actually needed for the job. Vision is much harder than sound.

 

[coliver988]

I haven‘t yet dived into this system, but I will note that these days voice recognition is typically done in the cloud, so there could be larger computers behind it. Not sure what is actually needed for the job. Vision is much harder than sound.

depends: Google new chip in theory does the recognition entirely on the phone from what I understand.

 

[Baroun Tardis]

OpenCV runs on single board computers and does object recognition without the cloud

Introduction to Computer Vision Using OpenCV and the Raspberry Pi

Learn how computer vision works by creating a deep-learning-based face detection project using OpenCV on a Raspberry Pi.

www.circuitbasics.com

Similar things exist for doing voice recognition without the cloud.

Recognizing Speech with a Raspberry Pi

Learn how to turn audio into text with the free, open-source spchcat tool on a Raspberry Pi. By Pete Warden.

www.hackster.io

============

Here's the house rule I'm going with:

There's "Robots/Bots" and there's "Droids"
Droids have real C4/C5/C6, can carry on conversations, and learn new skills through experience or improve skills. They use any kind of brain. The T5 robot rules produce "Droids" , all of them, as written.

Robot/Bots have C4<=4, C5=Instinct, can't learn new skills, have to be programmed. You can do file swapping, or cartridge programs, and change which skills are in play. They don't use personalities, they're coded expert system driven things that run only on electronic (photonic, etc) brains. Because there's not a whole personality there, they can't handle novel situations or complex tasks. A player with Programmer skill can write Python-ish code to string skill usages together, but the GM gets to be the one to decide how that executes based on the situation.
You can use things from CT Book 8 in building them as desired, so "Basic Sensor Package" for cr1700 is OK.

Example:

Helperbot

9

20kcr

7

The helper-bot has a wide plastic base with tank-treads underneath. The wide base is used as a shelf for delivering requested items. The sensor pod at the top allows it to look around, and the two small arms let it hand the user tools or other items requested. It’s smart enough to understand “follow me” or “Remember this <landmark name>” or “Go to <landmark> and get the wrench” There are four small compartments around the edge of the base, each with a tray inside that can extend on command to provide tools or other small bits of gear. (“Robot, extend the medical supply tray”) (“Robot, ammo tray open”) The robot has a short range communicator (5km or so) and a standard Comm built in, with Agent software (It can do the kind of things Google Assistant would do.) The black square on the chest can be opened to reveal a standard device charging port