Visualized using Foxglove: Unitree's Go1 Quadruped
Try it for yourself: https://foxglove.dev/examples 🔥
For those who haven’t met us live at a conference, we’d like to introduce Bodi, Foxglove’s unofficial “mascot” – a Unitree Go1 visualized using Foxglove. 🐾
In this visualization, we’re highlighting two key aspects:
1. We’re projecting images onto a spherical sector in 3D rather than a rectangular hyperplane, because, well, it looks awesome.
2. Since the stereo cameras on the Unitree Go1 are close together, and its point clouds are generated from these cameras, we chose to use a Stixel view with a decay time of 0.5. This compresses the 3D data into a quasi-2D format, which enables faster processing and decision-making, while preserving crucial spatial relationships between objects and the environment.
Stixel views provide an efficient way to handle point cloud data by reducing the complexity of raw 3D points. This approach segments the scene into vertical columns, or “stixels,” which represent planar regions like road surfaces, obstacles, or open space. This abstraction is computationally lighter than processing individual points, making it ideal for real-time applications. If you’ve visited us live at a conference, such as hashtag#ROSCon2024 in Odense, Denmark, you would have seen Bodi visualized live using the easy to use foxglove_bridge and websocket connection.
The stixel framework boosts performance in tasks like object detection, path planning, and obstacle avoidance, allowing robotics systems to operate more efficiently and with less computational load. Additionally, it integrates seamlessly with stereo vision and lidar sensors, improving robustness across diverse conditions.