---
title: Explore with a simulated ROSMASTER R2
description: Run a Python app that explores an indoor world using lidar, depth and odometry, with a live map and timed start/stop controls.
language: python
---

[R2 Explorer](https://github.com/wendylabsinc/WendyOS/tree/main/Examples/R2Explorer)
drives a simulated ROSMASTER R2 through an indoor world. It builds an observed map,
selects destinations it has not visited, and checks curved forward and reverse
paths against lidar returns. A front depth check limits forward movement. The
planner reads sensors and odometry, without reading the simulator's obstacle list.

This sample requires the managed `rosmaster-r2` VM profile. It has no physical
robot backend and refuses to start without the simulator's identity and sensors.

## Prepare the simulator

You need Git, Go, Docker, and the [Wendy VM prerequisites](../../../../installation/wendyos-virtual-machine/#what-youll-need).
Build Wendy from the same checkout as the sample so the CLI includes its R2 runtime:

```sh
git clone https://github.com/wendylabsinc/WendyOS.git
cd WendyOS
make -C go build-cli
./go/bin/wendy vm create r2-explorer --profile rosmaster-r2
./go/bin/wendy vm robot start r2-explorer
./go/bin/wendy vm robot open r2-explorer
```

Keep this terminal in the repository root. The first build needs internet access.
For an existing VM with an older agent, follow the
[R2 runtime setup](https://github.com/wendylabsinc/WendyOS/tree/main/go/simulator/rosmaster_r2#use-a-managed-wendy-simulator)
to update the agent. After rebuilding a CLI with changed simulator sources, use
`./go/bin/wendy vm robot update r2-explorer` to update the pinned runtime.

## Deploy and explore

```sh
./go/bin/wendy run --prefix Examples/R2Explorer --device vm:r2-explorer --detach
```

Open [the exploration dashboard](http://127.0.0.1:3510). It starts parked. Select
**Start exploring** to transfer app control and begin a run. The default duration
is 120 seconds; the dashboard accepts runs from 5 to 600 seconds. Watch the observed
map, driven path, goal and color/depth previews as the car moves.

Select **Stop** or press **Space** in the dashboard to stop early. Closing the tab
does not stop a run; its timer continues in the app container. The simulator's
Stop, Pause, Reset or controller selection also interrupts exploration. The app
stops on stale sensors, contact, loss of control or timer expiry and never resumes
automatically. Select Start again when ready for a new run.

Stop in the app sends zero motion for its own control session. The simulator's
Stop or controller selection revokes ownership. If another controller has taken
over, the explorer's Stop does not interrupt that controller.

Read the map and run state from a second terminal:

```sh
curl --fail http://127.0.0.1:3510/api/status
```

To stop the deployed app entirely:

```sh
./go/bin/wendy device apps stop henk-explorer --device vm:r2-explorer
```

The simulator's 300 ms command watchdog stops motion if the app disappears.

## Change the planner

`Examples/R2Explorer/planner.py` contains visit memory, goal selection and the
Ackermann path rollout. `app.py` handles sensor freshness, control ownership,
time limits and HTTP endpoints. The sample uses only Python's standard library.
Its unit tests and offline simulator acceptance check run from the sample directory:

```sh
cd Examples/R2Explorer
python3 -m unittest discover -s tests -v
python3 tests/simulator_acceptance.py
```

This is reactive exploration with ideal simulator odometry. It can revisit areas
or stop where no safe path is available; it does not implement SLAM or guarantee
complete coverage. Camera optics, depth noise and drivetrain dynamics are
idealized. A hardware port needs calibrated sensors, localization and a physical
motion backend. See [sim to real](../../../../integrations/sim-to-real/#using-the-rosmaster-r2-template)
for the supported interfaces and explicit service backend selection.