Authored or co-authored seven patents in autonomous-system control and human–AI collaboration. All center around making complex systems easier to use.
Setting up a fleet of warehouse robots means filling in a lot of technical settings, and one wrong value can send a robot somewhere it shouldn't go. This invention takes those raw configuration details and automatically rewrites them as ordinary sentences a person can read.
Most warehouse systems make you decide everything up front — where the robot is going, which route it takes, what it does when it arrives — before it can move at all. Real operations don't work that way: a truck shows up and the details arrive as you go. This invention lets a robot start a job knowing only part of the picture, travel as far as that information allows, and pick up the missing pieces from a person or another system along the way. If something still isn't known, the robot waits at that point and carries on the moment the answer arrives.
When a company rolls robots out to a new warehouse, someone usually rebuilds every workflow from scratch, even when the new site does the same work as the last one. This invention packages a proven workflow as a reusable template that captures the shape of the job — pick this up, then deliver it there — while leaving out anything specific to one building. The template can be exported, handed to another facility, and filled in with that site's own locations. Setup stops being a rebuild and becomes an adaptation of something that already works.
This is the foundation the others build on: a small vocabulary for describing any material-flow job out of four pieces — where to pick up, where to drop off, which location, and which route. Each piece is marked as either known ahead of time or decided in the moment, and those combinations compose into thousands of different workflows without anyone writing custom software. An operator can say “pick from any of these five spots, drop at any of those five,” and the system works out the possibilities and lets the robot choose as conditions change. It turns warehouse automation from a bespoke build into something that can be reused across sites.
When several robots want the same dock or drop-off point at once, they pile up and block the aisle. This invention gives every work location an associated waiting area, so a robot that finds its destination occupied is sent to queue out of the way and is let in as soon as the spot frees up. It also draws those waiting areas on screen alongside charging stations and service bays, so a supervisor can see where every robot is when it isn't actively working. Traffic stops being a mystery and becomes something you can look at.
Describing robot behavior traditionally meant stringing together dozens of interlocking if-then rules — effectively programming, and easy to get wrong. This invention reduces a job to two repeated questions: go here, and do this. An operator picks a location or a group of locations, picks an action such as pick, drop or wait, and chains those steps into a whole workflow, optionally with a trigger that starts it automatically. A facility manager with no programming background can build and change robot workflows themselves.
In most warehouses, robot performance and human performance are measured in completely separate systems, which makes work that people and robots do together very hard to improve. This invention tracks both in one place, collecting data from the robots and from electronically tracked workers as a job unfolds. It reads that data as it arrives and adjusts task assignments and routes on the fly. The point is to treat a mixed human-robot operation as one system rather than two that happen to share a floor.