June 9, 2021
If I missed something, click the "Edit" button and add it in! -Will
Above: Pieces needed for Demo 1. We've crossed out two whole blocks, and other pieces have made significant advancements, too! Compare with last week's chart.
- Working to install Lanelet2 painting plugins for JOSM
- Continuing to decode Autoware's strange formatting differences from vanilla .osm files
- Contacted maintainer of Lanelet2 repository
- Once a Lanelet map of campus is built, dynamic route planning should work seamlessly
- Nothing this week, but might have something to share soon
- Looked into Autoware's existing param and source files for clues with the MPC controller, but no luck so far (the controller is still publishing steering commands of "0")
- Explained a bit about how Model Predictive Control (MPC) works, and offered that the time horizon of 100ms might be too small
- Modeled a mount for our steering angle sensor. The mount has been 3D printed and successfully installed.
- Only two parts remain: the two gears to connect the steeting column to the potentiometer (sensor)-- Vishvak has offered to model these, too. Details are available here.
- Also offered to work on URDF once his schedule clears up. Thanks, Vishvak!
Above: Vishvak's new mount (black) holding up our potentiometer. Vishvak is going to model two gears to link the potentiometer and the shaft.
Above: Vishvak's mount finished after almost 8 hours of printing.
- Finalized steering interface and CAN code-- ready to go! 🎉
- Also wrote voltron_test_utils, which can help make our safe and stable
- Josh's finished code is the link between Cristian's MPC controller and the actual steering hardware. (See chart at top.)
- Working to align front and rear Lidar sensors with the vehicle origin (defined as the midpoint of the rear axle) so that Lidar fusion works.
- By the way, this involves a strange mathematical concept called quaternions, which Will learned a tiny bit of this week thanks to the internet.
- Once Lidar fusion works, we can create a 3D map of campus.
- Mostly tabling Demo 0 (see last week), focusing on real-world sensor calibration for now.
- Having to rebuild VDE on the Jetson to add necessary dependencies for Demo 0. This is a pain and it takes ages, but it should smooth things significantly.
Above: Our front and rear lidars are currently misaligned, and you can see that in RViz. Are we in two rooms at the same time?
- 38 days until the DFW AI Self-Driving Small Vehicle Showcase