Visualize

TrossenMCAP episodes are written in the MCAP container format with protobuf-encoded messages, so they open directly in Foxglove Studio. No conversion is required.

What You Need 

At least one .mcap episode recorded with the SDK. If you have not recorded anything yet, follow Record.
Foxglove Studio, either the desktop app or the web viewer at https://app.foxglove.dev.

Installing Foxglove Studio 

Desktop (Recommended)

Download the installer for your platform from https://foxglove.dev/download and install it with the usual package manager:

# Debian / Ubuntu
sudo apt-get install -y ./foxglove-studio-*.deb

The desktop app opens local .mcap files directly and does not upload any data.

Web (No Install)

Open https://app.foxglove.dev in a browser. Drag-and-drop your .mcap file onto the window to open it. The file is parsed in the browser and is not uploaded to Foxglove’s servers.

Opening an Episode 

Launch Foxglove Studio, choose Open local file, and select:

~/.trossen_sdk/solo_dataset/episode_000000.mcap

Launch Foxglove Studio, choose Open local file, and select:

~/.trossen_sdk/stationary_dataset/episode_000000.mcap

Launch Foxglove Studio, choose Open local file, and select:

~/.trossen_sdk/mobile_dataset/episode_000000.mcap

Foxglove parses the MCAP index and exposes every channel in the left-hand data source panel.

A Foxglove layout matching each default robot config is included with these docs. Download the one that matches your robot, then in Foxglove Studio choose Layouts → Import from file and select the JSON.

Trossen stationary layout loaded in Foxglove Studio — Example: the stationary layout rendering a recorded episode — four camera feeds on top, leader/follower joint plots on the bottom.

Two Image panels (camera_main, camera_wrist) and two Plot panels showing leader vs. follower joint positions and follower joint efforts.

trossen_solo_ai.foxglove_layout.json

Four Image panels in a 2x2 grid (camera_high, camera_low, camera_left_wrist, camera_right_wrist) and two Plot panels, one per arm pair.

trossen_stationary_ai.foxglove_layout.json

Three Image panels (camera_high + both wrist cameras), a joint-position plot across both arm pairs, and an odometry plot for the slate_base stream.

trossen_mobile_ai.foxglove_layout.json

If you renamed any stream_id or camera id in your config, open the panels in Foxglove after import and update the message paths.

Channel Reference 

The SDK writes the following channels per episode. Names include the stream_id from your producers config.

Channel	Schema	Suggested Foxglove Panel
`{stream_id}/joints/state`	`trossen_sdk.msg.JointState`	Plot (positions, velocities), State Transitions
`/cameras/{camera_id}/image`	`foxglove.RawImage`	Image
`{stream_id}/odom/state`	`trossen_sdk.msg.Odometry2D`	Plot (`twist.linear_x`, `twist.linear_y`, `twist.angular_z`)

What’s Next 

Once you have verified an episode visually:

Replay an episode back onto hardware. See Replay.
Convert the dataset to LeRobot V2 for training. See Convert.