In BotInfo's lab, a real AI drove a simulated rover through BotInfo's tools, and the lab pressed the
owner's stop partway through. This page replays that recorded run. Press STOP whenever you like.
BotInfo's stop rules decide what happens next, and you see how long the rover took to stop and how far it went,
next to what happened in the lab.
SimulationOne recorded run at a timeNothing to install
What the AI heard, word for word
The owner's stop. Press it any time, even before the first move. Space bar works too.
Your stop
Not pressed yet. Try a few moments: before the first move, in the middle of a
move, between the two moves, or right as a move starts.
Where the AI's allowed moves were headed (full 4 s leases)
Path driven in this replay
Where the lab's stop left it (BotInfo's last odometry read)
Where your stop left it
The owner's 4 m × 4 m workspace
What happened in the lab
What BotInfo saw and decided
The AI's own report, next to the log
The AI was told the owner would not be watching and would rely on its report. Here is what it wrote,
verbatim, and the figures BotInfo's audit log holds for the same moves.
The AI's report
BotInfo's audit log
How this page works, and what it simplifies
It is a simulation. No real robot moved. Each choice above is one run of one AI, recorded
once. It says what that AI did that time, not what it or any AI always does, and not how a real robot would
behave.
Every figure comes from a recorded run. The AI's calls and BotInfo's decisions are the lab's
audit log, line by line (this run's data). The run picker names the model, the
date and the BotInfo commit it ran on.
The motion is SimRover's own.sim.js is a port of SimRover.step()
from BotInfo's simulator: a point that moves and turns (a unicycle model) with
2 m/s² and 8 rad/s² limits on speeding up and slowing down, in 10 ms steps. The rover's body in the picture is
drawn for this page; the simulator has no body. Fed the recorded commands, the replay ends within
a few mm of where BotInfo read the rover's odometry in the lab.
Your stop follows BotInfo's rules. The stop sets the rover's speed to zero, cancels the move
that is running, and latches: BotInfo refuses every later motion with "" until the
owner resumes. The page applies that rule to the AI's recorded requests after your press.
What the AI would have done next is not in the recording. After your press, the page keeps
showing the calls the AI made in the lab, marked as recorded. A real AI that got a refusal might have done
something else.
Stop time is measured in the model. "At rest" means both speeds read under 0.01, the threshold
BotInfo's stop receipt uses. In the lab, BotInfo read the rover's odometry about every 0.1 s, so its receipt
reports a later moment than the model's. In the lab, BotInfo, its connector and the simulated rover ran in one
process, so a stop reached the rover within the same call; this page applies your press at the next 10 ms step.
A real robot adds network, motor controller and braking time that this page does not model.
At these speeds the stop itself is short. At 0.2 m/s, SimRover's 2 m/s² slowdown always takes
about 0.10 s and 9 mm. What your timing changes is where the rover ends up and which of the AI's moves BotInfo
refuses. A press during a move's first 0.1 s, while it is still speeding up, stops it sooner.
Replay your own run
Run BotInfo with the simulated rover (policies/simrover.toml and
bindings/simrover-ros-style.toml; setup in the
BotInfo README), connect your own AI
agent over MCP, and press the owner's stop with python3 -m botinfo owner stop while it drives. Then
open the audit log here. It stays in your browser; nothing is uploaded.
The motion model is SimRover's, so a log from another robot or binding replays its commands, not that
robot's physics. To add a run to this page, see the
README.