FRC 6328 Mechanical Advantage 2023 Build Thread

[jonahb55]

@jonahb55

This project began in fall of 2022, and we have a couple of reasons for taking this route rather than using a prebuilt system like PhotonVision or Limelight .

• We used this project during the offseason as a training opportunity, bringing in some of our newest members. We were fortunate enough to have the time and resources to take this extra step without making sacrifices in other areas.
• A custom solution gives us a greater ability to optimize for our specific choice of hardware, both in terms of performance and reliability. We can quickly debug any issues we encounter or experiment with different approaches to improve performance.

On the second point, we were able to achieve better performance than PhotonVision using OpenCV’s Aruco module. At the time, PhotonVision only supported the AprilTag3 detector. PhotonVision now supports Aruco too, but I don’t know a lot about it or what kind of results they’re getting. On different hardware, some people have observed poor accuracy using Aruco but we had no such issues (resolution may be a factor here). Limelight is the other option of course, but even the LL3 has relatively poor AprilTag performance and no global shutter option.

The other benefit was that a custom solution made it much easier to debug issues, including subtle issues with the camera and network connections. The last time we tried PhotonVision was in 2022 and we had numerous issues (most of which have since been fixed). Given how important the vision system was on the robot, we couldn’t risk having an issue in the middle of competition that we couldn’t debug ourselves. With a custom solution, we only needed to support one set of hardware and we controlled the whole stack. I wouldn’t recommend this approach in general though given that PhotonVision is more reliable now and has better community support.

We also added some extra features that weren’t available elsewhere. The whole system is configurable over NT (which is way nicer trying to use a web interface and keep settings in sync).

@jonahb55

Our plan is to have four cameras on the robot processed by four Northstar instances running on multiple coprocessors. Given this requirement, we wanted to make it easy to configure and manage multiple instances. A traditional web interface makes it challenging to keep track of each device’s configuration. Instead, Northstar relies on NetworkTables as its primary interface.

An MJPEG stream is available through a browser when necessary. All of the camera settings can be tuned on the fly, but the real benefit of using NT is that the default configuration is stored by the robot code and pushed to the coprocessors on startup. After retuning one camera, we just adjust the constant, restart the code, and every instance will be using the correct configuration.

We also used a ChArUco board instead of a checkerboard for calibration, which has since been adopted by Limelight and PhotonVision (I think?).

@jonahb55

Northstar uses a ChArUco board for calibration instead of the traditional checkerboard. This has the advantage of not requiring the full board to be visible, making calibration much more robust to disturbances like odd lighting (or a hand covering the board).

Northstar will continue to improve over time too. There are several new features the team is working on for 2024, so there are no plans to switch to another solution in the near future.