I’ve been noodling with some custom python scripts for the pidog. I have quite a few now, thought I’d share with anyone who is interested. Some of them are just proof-of-concept, so I’d really love some help fleshing them out. Even if it’s just testing and giving some feedback.
I’ve put a description of each module, and my future plans for tweaks and upgrade.
The one I consider most complete is Voice Patrol. This runs a patrol routine that makes the pidog wander, map its environment, and avoid obstacles dynamically. It also has voice control with a growing list of commands, active while the pidog patrols. Voice command proved a little tricky to set up, and I’ve yet to write a proper installation guide for how I got it to work, but stay tuned. For now, the other patrol modules offer the same functionality without voice commands, all with different approaches to navigation and obstacle detection.
There are other modules for facial and object recognition, still very much in the early stages.
Anyone who wants to, give them a look and tell me what you think. If you have any suggestions for other modules, please let me know and I’ll add them, or you can push your own commit if you like. My ultimate goal is to eventually have a master script that imports and calls the other modules at need, for a truly autonomous and full-featured pidog.
You can refer to PiDog’s GPT example where it patrols while listening for voice commands. Upon detecting an instruction, it stops, performs the corresponding action and response, then resumes patrolling.
Additionally, to avoid interference from servo noise, you may need to adjust the sound detection threshold.
Brilliant, thanks for sharing.
I shared a few of my own code and scripts on this forum, for voice control etc
Demonstrated here..
As per your own software, the joystick control demo uses the IMU for gait adjustment for stability.
And autonomous behaviour using orbslam3 + AI object/hazard recognition, These data are fused with the camera extrinsic matrix to create a real world (scaled) occupancy map. E.g green safe floor, red = hazard
Awesome! I love your work. I’m not familiar with ROS at all but I’ll take a look at this today. Vision-based SLAM is on my wishlist, thanks for sharing.
Your chat gpt method requires paying for openAI credits lol. Much more satisfying to find a free method, or host your own chatbot. I don’t understand why the only method you show in documentation for AI integration is also the most expensive option out there.