5 open source tools compared. Sorted by stars. Scroll down for our analysis.
By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.
| Tool | Stars | Velocity | Score |
|---|---|---|---|
ardupilot ArduPlane, ArduCopter, ArduRover, ArduSub source | 15.8k | +46/wk | 81 |
Rerun Visualize, query, and stream to train on multimodal robotics data. | 11.4k | +48/wk | 80 |
PlotJuggler The Time Series Visualization Tool that you deserve. | 6.1k | +10/wk | 62 |
Lichtblick Lichtblick is an integrated visualization and diagnosis tool for robotics, available in your browser or as a desktop app on Linux, Windows, and macOS. | 1.1k | +15/wk | 58 |
Dalaran Dalaran: Apache-2.0, robotics-first visualization and data infrastructure for multimodal time-series. ROS 2 native, reads existing .rrd recordings. A hard fork of Rerun. | 795 | - | 61 |
Stay ahead of the category
New tools and momentum shifts, every Wednesday.
ArduPilot is the autopilot software behind over a million drones, rovers, boats, and submarines. The entire stack is GPLv3, no paid tiers, no cloud lock-in. You flash it onto a Pixhawk flight controller and get GPS navigation, mission planning, autonomous waypoints, return-to-home, and failsafe handling out of the box. NASA, Boeing, and Intel all use it in production vehicles. Running ArduPilot means running real hardware. You pick a flight controller (Pixhawk is the standard), flash the firmware, calibrate sensors, and configure your vehicle type. The toolchain is C++ and Python. Ground control stations like Mission Planner or QGroundControl handle the UI side. Expect a weekend to get your first vehicle flying reliably, longer if you're building custom frames. Solo hobbyists and university research labs get everything for free. Commercial drone operators use it to avoid DJI's ecosystem lock-in. Large teams building autonomous fleets use ArduPilot as the base layer and add their own control software on top. The catch: this is aerospace firmware, not a web app. Misconfiguration can crash a physical vehicle. The GPLv3 license means any firmware derivative must also be open source, which some commercial OEMs avoid.
Rerun is a visual debugger for anything with sensors. Log images, point clouds, 3D transforms, joint states, and time series from Python, Rust, or C++, and the viewer replays them on one shared timeline so you can scrub to the exact frame where the robot decided a wall was a door. pip install rerun-sdk is the whole install, and it is free under Apache-2.0 and MIT. There is nothing to operate. The viewer is a desktop binary or a browser tab, recordings are .rrd files on disk, and the query layer runs locally over columnar storage with SQL and dataframe APIs. It reads MCAP, so existing robot logs load without rewriting your instrumentation. Solo researchers and small robotics teams get the whole thing for nothing, and the free tier is the actual product rather than a trial. Larger teams needing a shared catalog, dataset streaming to GPUs, SSO, and single-tenant hosting are looking at Rerun Hub, quoted per deployment with no published price. The catch is churn. The project says outright to expect breaking changes, the viewer bogs down with too many entities or multi-million point clouds, and a hard fork already exists: Flaminis/Dalaran is ROS 2 native and reads the same .rrd recordings. Pin your version.
PlotJuggler plots time series fast, and it is what robotics and vehicle engineers reach for when something went wrong and the answer is buried somewhere in a log. Drag a signal onto the canvas and it draws. Drag another and they overlay. Scrub the timeline and every plot moves together. MPL-2.0 and free. It is a desktop application, C++ and Qt, installed from a package, an AppImage, or a snap. Nothing to host and no account anywhere. It reads ROS 1 and ROS 2 bags, MCAP, CSV, and ULog, and it can stream live from a running system rather than only from files. The transform functions are the feature people stay for: derived signals get computed inside the tool instead of sending you back to a Python script. Solo engineers debugging robot or drone logs: install it, it costs nothing and repays the effort the first time you need to line up two signals that disagree. Teams: free, and it pairs naturally with lichtblick-suite/lichtblick, which covers the 3D scenes and camera panels that PlotJuggler deliberately leaves alone. The catch: PlotJuggler plots numbers and nothing else. No 3D view, no image panels, no point clouds. It covers one pillar of what a tool like Foxglove does rather than replacing it. The interface is also unapologetically an engineering tool, dense and thoroughly Qt flavored, and it is not going to win any design awards.
Lichtblick is what happened after Foxglove closed its studio. Bosch forked the last open source version and kept building on it. Load a robotics log, MCAP or ROS 1 and ROS 2 bags, and get 3D scenes, plots, image panels, and diagnostics in a layout you arrange yourself. Runs in a browser or as a desktop app on Linux, macOS, and Windows. MPL-2.0. Nothing to host. It reads files on your machine or connects live to a running robot. Development is active with releases most months, and recent ones added H.265 video playback, a text filter in the node editor, and a rebuilt alerts system. The extension API lets you write panels for message types nobody else has, which is most of them once your robot has custom messages. For teams who want the Foxglove Studio experience without an account or per seat pricing, this is the direct answer rather than an approximation. Solo researchers and students: free and complete. Small teams: free, and the logs never leave your machine, which matters when they contain customer sites. Large teams: free for the visualization layer, but Foxglove's hosted data platform, the part that searches across a whole fleet's worth of recordings, has no equivalent here. The catch: this is a fork maintained by one company's team rather than a broad independent community, and it inherits a large Electron and TypeScript codebase. It tracks what it forked from more than it races ahead, so when Foxglove ships something genuinely new in the closed product, you either wait or you build it yourself.
Dalaran is robotics visualization that speaks ROS 2 natively. Log topics, LiDAR, camera streams, and transforms, then get synchronized 3D playback and dataframe export. It's built on Rerun's bones, Apache 2.0 all the way down, and it opens existing Rerun recordings without conversion, so you can point it at your current data today. Install is pip, cargo, or CMake, with a doctor command for the inevitable ROS 2 environment wiring. What it adds over Rerun is the domain layer: joint states, URDF-driven transforms, correct nav2 costmap semantics, things you'd otherwise hand-marshal. It even ships the self-hostable catalog server, the piece Rerun's commercial strategy sits on. ROS 2 teams who found Rerun's generic time-series model didn't speak robotics are the audience. Foxglove is the proprietary incumbent; RViz is the free classic that never left. The catch: this is one person maintaining infrastructure that competes with two venture-backed companies, and the README warns that breaking changes are expected between releases. Impressive work with a real bus-factor question attached.