Machine builders: would you trade mechanical determinism for robot + vision adaptability?
retained text
I’m curious how OEMs and machine builders here see this. When a machine has to handle many product formats, we usually force determinism mechanically: guides, feeders, nests, format parts, adjustments, etc. I’m experimenting with the opposite idea: keep the mechanics simpler and push more variability into robot + vision + software. The video is a small example: randomly oriented caps are detected in 3D, picked with suction, corrected with a second vision step and placed onto a bottle. The next step I’m interested in is going beyond deterministic robot logic and using trained policies / Physical AI for bounded industrial tasks, while keeping PLC and safety layers deterministic. So where would you draw the line? A) more mechanical complexity, but predictable behaviour B) simpler mechanics, with robot/vision/software handling more variability And if you’re an OEM or machine builder: what would make you reject B? Cycle time? Reliability? Validation? Serviceability? Debugging?
document fetched 2026-10-12 00:33 · canonical https://reddit.com/r/robotics/comments/1x3601w/machine_builders_would_you_trade_mechanical