Resolved by DJ Sures!
If you have driven any modern car they have a lane keep assist function. Essentially it will just follow the lines on the road and stay in the lines. My robot works quite well with NMS in enclosed areas and avoids objects etc but when it gets out in the open on the footpath you have to manually steer it to move between locations. It would be really good if it would just look at the foot path and then keep following it without going on the grass or driveways.
That said what would the best way to go about trying to make the robot follow the footpath and walk autonomously until corrected? The path for the most part is the same off white colour so I thought about just colour tracking but that always isn't the case and many people have concrete driveways. Open to ideas.
Most robots that work in warehouses or hotels, etc have some external markings or floor transmitters. The markings would be easiest to implement.
perhaps lines among the outer edge? Or simply down the middle of the walkway
The Use case is a delivery robot and being able to control several at once so I think it is going to have to be a neural network machine learning type of approach based on learning from manual control over time. It’s easy to control one robot one camera one joystick at a time but when you have 1 person trying to control 30 of them .
played for a while today down my street (neighbours all give me strange looks these days). Not having a lot of success. Wondering if something like this is on the roadmap for Exosphere?
I found this earlier and it may be of interest. It's a hack to make a self driving 1/10scale car. The guy built it to drive around the sidewalks of his neighborhood.
https://medium.com/@ericmaggard
Thanks I will close this. I do look forward to the AI and ML coming to exosphere in the future as I am sure this would help train the robot every time we remote control it so it understands what is footpath and what is not.