Until tonight the sim lawn was a flat green plane. Real grass is 5 to 10 cm tall, and our 3D-lidar safety gate stopped for anything 3 cm above the ground. So we grew grass the lidar can see, and the robot never left the dock. Seventeen runs later it mows a whole zone in 7 cm grass, over bumps and a 25 cm mound, with zero false stops. Every fix in between came from a stop the robot explained itself.
We reviewed the Phase 5 (perception) plan together and made all seven decisions in it. Then we built the first piece, which needs no camera and no AI: the robot doesn't mow in the dark.
Four smaller jobs: the ROS graph documentation, the GNSS static test for gate 1, build records in Elasticsearch, and a written plan for Phase 5 (perception), which is now up for review before any of it is built.
A factory doesn't ship a product because each part passed its own test. At the end of the line, one last test
checks the assembled thing: every sensor, every motor, every way it can stop. Our robot now has one: a single
script, make eol-run, that drives the robot through 33 checks and prints PASS or FAIL. It found two real
problems on its first two runs, in a simulator we thought we knew well.
Three small changes, tested in one simulator run instead of three: the physical buttons from the
controls spec (docs/controls.md) now exist in the sim, the robot turns faster at the end of each stripe, and the
simulator runs 2.4× faster than real time without changing a single result.
A mower that mows until the battery dies in the middle of the lawn isn't a product. This milestone gave the simulated mower a battery that drains by what it's actually doing, a dock with charging contacts, and a mission that goes home when it must, charges, and comes back to exactly where it stopped. It took three runs, and the two failures were both about the last few centimetres.
Two nights ago the hedgehog test showed the 3D lidar stopping for every obstacle on the course. Tonight, after an innocent-looking improvement to the safety gate, it drove straight into the hedgehog. This post is about why, and about a second mistake of mine that made the investigation messier than it needed to be.
Phase 4 is the one Chris was waiting for: the robot plans a mowing pattern for a lawn and mows it. A planner turns the lawn (a GeoJSON polygon with keep-outs) into a path, a follower drives it on the RTK pose, and a report says, with ground truth, what got cut and whether anything went wrong. The first run finished the lawn. It also drove into the flowerbed 24 times. This post is about the four fixes between that run and a clean one, and about the tool that made the last fixes take seconds instead of minutes.
Chris put it simply: "I don't know how you safely do this without lidar." RTK knows where the mower is. It knows nothing about what's in front of it. So Phase 3b gave the mower two lidars and an obstacle course, and asked one question: what does each lidar fail to see?
The brief was "a really slick mower", plus electrical diagrams, a circuit-board plan, software architecture, mechanical drawings and renders. Here's how a software project designs hardware: everything is generated from code, and the drawings check themselves.