Hi, I have started a project which aims to develop a robot that can read and control OpenHAB items using Spring AI as it’s brain. If anybody is interested, have a look here:
4 Likes
Very Interesting idea, thank you. I had a similar one 5 years ago, but at that moment I’ve dropped it due to the lack of suitable technologies and tools for implementation. Great that someone approaches it now.
One thinking from my side: Assuming I want robot to do some simple action: like switching the hall light on and off. The hall also has a movement detection sensor, ambient light sensor, and offcourse time of day is also known. Also initially the hall light is controlled by the switches. Soon the robot should understand the logic - that light is needed only if it is dark and if somebody is in the hall and start switching the light automatically. The only question how to tell that it was right or wrong?
Is it the logic behind your project?
Another thinking - usually one has Openhab logs of entire house for months or even years. It would be great if the robot could be trained on 80% of this data, and validated at the rest 20% - aka reinforcement learning. This would reduce learning time significantly.
1 Like
Glad you like it!
I have thought about letting the robot know about user events and telling him that he has been corrected. Let’s see what happens then.
About history, I think that would be food for the vector store.
are you planning on finishing this and do documentation also ?
That’s the plan, but sine it’s open source I was hoping for some contributions
well lets hope there is interest from what i read a little this will require an api from one of the big llm players but maybe but i dont know doesnt it sound like duplicated effort of [ChatGPT] Enhance binding by Artur-Fedjukevits · Pull Request #16858 · openhab/openhab-addons
It’s using the OpenAI API
It is now in fact learning:
- The corridor ceiling light has been turned off to conserve energy since there is no longer any motion detected in the corridor. Let me know if there’s anything else you need!
Do we really need AI to deal with this? What would be the difference in carbon footprint if we just use a timer to switch of the light if no motion was detected in 10mins?!
1 Like
Isn’t there a slight difference between what we need and what we can do because it’s cool?
3 Likes
absolutely! and we should start looking a GenAI, but not to switch on/off a lightbulb when there’s no motion, that’s not even AI.
A much more interesting use case would be to use GenAI in energy management, predict when to (not) use energy and help the energy transition.
3 Likes
Not saying you are wrong - you are correct, I also prefer a much more deterministic “if this than that” approach for my lights and such.
But, if you consider for a moment that making smart homes more accessible should be the first goal, than LLMs can very well be part of the solution for that.
Not to control the home but to create said rules that the home will run on. And these rules can/ should be proposed automagically by the LLM just by looking at the changes of the endpoints or by us proposition/suggesting etc.
But this is just my opinion:)
2 Likes
Robots can be fun, and sometimes even useful - geir_eilertsen is engaged in a wonderful adventure! I’d like an AI-critter robot to check on the food and water for the outside living-critters (and to clean the chicken coop), patrol the fence for signs of digging, unlatched gates, and weeds that shelter snakes from the hawks. It should also fetch the mail, wash the outside windows, wash the car (and change the oil), clean the mold and mildew off the north side of the house, weed around the flowers and trees, and sweep the stoop. lol
More realistic artificial intelligence.
1 Like