Release 1.0 Task List

[Levi-Armstrong]

• Leverage boost_plugin_loader Update to leverage boost_plugin_loader repository #1094
  • Need to port this functionality to boost_plugin_loader. More details in this PR Static plugin loading using symbol module resolution #782
• Add floating joint support Add floating joint support #1060
• Add mimic joint support
• Add planar joint support Draft: Add basic planar joint support in the urdf parser #915
• Leverage tesseract xml namespace in urdf parser Update URDF parser to use Tesseract XML namespace #1081
• Move profile dictionary class to tesseract_common
• Migrate profile to plugins and leverage yaml file
  • Possibly add reflection capabilities (reflect-cpp, rttr)
• Descartes Improvements shown here
• Investigate improvements to task composer ports
• Investigate improvements to profile dictionary to avoid needing static key function
• Investigate updating task composer and data storage so every graph and sub graph gets an isolated data storage
• Investigate improving logging functionality within the task composer
• Investigate removing/replacing console bridge. [This may be a possible starting point]
  • I believe leveraging spdlog might be the best option available
• Finish SimplePlanner profiles here
• Refactor everything still called PlanProfile to MoveProfile, see here.
• Add a tool for parsing task composer yaml and producing dot graphs
• Improve code coverage such that all packages are > 90%
  • tesseract_common
  • tesseract_collision
  • tesseract_environment
  • tesseract_geometry
  • tesseract_kinematics
  • tesseract_scene_graph
  • tesseract_srdf
  • tesseract_state_solver
  • tesseract_urdf
  • tesseract_command_language
  • tesseract_motion_planners
  • tesseract_task_composer
  • tesseract_time_parameterization
• Finish or clean up the very old Trajopt and Trajopt-Ifopt TODO lists.

Component updates

• Update TOTG to match MoveIt(2), some discussion here.
• Port Ruckig changes, here.

Support

• Define cmake target naming convention and update all packages
• ReadTheDocs Documentation
• CI Debian release process
• Prepare for and start respecting Semver once we're at 1.0, and add CI tooling to check this.
• Clean up all outdated issues, discussions and PRs.

Trajopt-Ifopt

• Update Trajopt-Ifopt to leverage Node Variable Set
• Finish the weighted sum logic in Trajopt to match Trajopt-Ifopt to validate the (new) approach of the latter, here.
• Update the osqp-eigen interface. Currently 0.6.3 is specified here, with the interface requiring work to support the latest osqp-eigen here. Update OSQP dependency to v1.0.0 and OSQPEigen to v0.10.3 trajopt#474, Update OSQP dependency to v1.0.0 and OSQPEigen to v0.10.3 tesseract_planning#664

Wishlist

• Coal (formerly HPP-FCL) as a discrete collision checker, here.
• Coal as a continuous collision checker (not started yet).
• Add toleranced Cartesian waypoints to Trajopt-Ifopt, here.
• Add a GUI for configuring Task Composer Pipelines and Profiles

Relevant Discussions

• Update namespaces to follow this convention tesseract::collision, tesseract::kinematics, etc.
• Make Tesseract a single CMake package instead of individual cmake packages.

[robinsonmm]

@marrts see this discussion. Open to ideas, and comments on things like Descartes etc

[johnwason]

RPI has also developed a very fast inverse kinematic solver IK-GEO: https://github.com/rpiRobotics/ik-geo

I had planned to work on this end of last year but may not get to it until the spring.

@aelias36 @controlguy168

[rjoomen]

RPI has also developed a very fast inverse kinematic solver IK-GEO: https://github.com/rpiRobotics/ik-geo

I had planned to work on this end of last year but may not get to it until the spring.

@aelias36 @controlguy168

This IK-solver might also be of interest: https://github.com/steffanlloyd/quik

[aelias36]

RPI has also developed a very fast inverse kinematic solver IK-GEO: https://github.com/rpiRobotics/ik-geo

I had planned to work on this end of last year but may not get to it until the spring.

@aelias36 @controlguy168

Here are some features of IK-Geo which I hope are of interest:

• Very fast. For example, 40x faster than IK-Fast for the UR5.
• Returns all solutions, not just solutions close to an initial guess.
• Robust to internal and boundary singularities. Continuous approximate solutions means that joint values are returned even when the end effector is right at the workspace boundary or even slightly outside it.
• Compatible with not just 6-DOF robots but also 7-DOF robots (using shoulder-elbow-wrist angle) and parallel robots.
• Compact source code makes it easy to port and modify. Many people have already forked it and provided improvements.

IK-Geo repo: https://github.com/rpiRobotics/ik-geo
IK-Geo paper (preprint): https://arxiv.org/pdf/2211.05737
7-DOF paper: https://www.sciencedirect.com/science/article/pii/S0094114X24002519

[marrts]

I think an extremely useful thing to add, I don't think it strictly needs to be done for 1.0, would be better logging control for everything that is output to the terminal currently. This way users have more control over what goes to the terminal and unique task output could be mapped to individual tasks, which could then be saved to files later for easier debugging, especially when running tasks in parallel.

Also, the current console_bridge use can be slightly annoying. The character count is limited and it locks a common thread every time it is called, which likely has a negligable effect but maybe could be better.

I don't think something needs to be built completely custom for tesseract from the ground up, but I think there are opportunities for some quality of life improvements there.

[johnwason]

We should also enable pre-commit before a 1.0 release. I started work on this in #996.

[johnwason]

Here are some notes on ABI compatibility policies:

• Debian policy on versions and ABI: https://www.debian.org/doc/manuals/debian-kernel-handbook/ch-versions.html
• ROS ABI policy: https://www.ros.org/reps/rep-0009.html
• ROS quality categories containing ABI policies: https://ros.org/reps/rep-2004.html

ROS seems to generally follow the Debian release policy. The "ROS Rolling" distribution is similar to "Debian Sid": https://wiki.debian.org/DebianUnstable

[Levi-Armstrong]

Related to logging, I would like to add abseil-cpp as a depend and start levering some of its features like logging. Probably need to make that call prior to a 1.0 release.

[johnwason]

@Levi-Armstrong it would be helpful to have a way to associate log messages with a specific taskflow instance and/or step.