Skip to content

Pose2d Command & Obstacles - #30

Merged
jgillick merged 20 commits into
v1.0from
obstacle_avoid_robot
Sep 7, 2026
Merged

jgillick merged 20 commits into
v1.0from
obstacle_avoid_robot

Conversation

@jgillick

@jgillick jgillick commented Aug 30, 2026

Copy link
Copy Markdown
Owner

Added:

  • A robot with ultrasonic range sensors
  • Position manager
  • Rewards for target position
  • Add support for list of entities to contact manager
  • Action managers can drive several joints from one action
  • NOTE: Fixed a bug to clear previous step values (observations, terminations, actions, etc) on reset. Your environment might need to enable obs_normalization=True if they stop reaching convergence after this change.
  • Fix: Observation noise is applied before scaling

@jgillick
jgillick changed the base branch from main to angular-velocity-indicator August 30, 2026 21:18
Base automatically changed from angular-velocity-indicator to v1.0 September 5, 2026 00:24
@jgillick
jgillick force-pushed the obstacle_avoid_robot branch from c8af99b to aa82f90 Compare September 5, 2026 00:48
@jgillick
jgillick marked this pull request as ready for review September 6, 2026 18:57
@jgillick jgillick changed the title Obstacle avoidant robot Pose2d Command & Obstacles Sep 6, 2026
@jgillick
jgillick merged commit 8e476c5 into v1.0 Sep 7, 2026
1 check passed
jgillick added a commit that referenced this pull request Sep 15, 2026
* Convert mdp class functions to data dicts

* Use config item execute method for all managers

* Remove deprecated parameters

* Video: in-process videos are named ".recording.mp4"

* Fix range setting bug

* Unbound get_terrain_height

* Fix misc bugs

* Add test coverage

* Update upgrade doc

* Add testing github action

* Add rsl-rl to the genesis dependency list

* Update uv lock

* Update test coverage

* Update example environments

* Add required mdp function params

* Lint all code

* Fix tests

* Remove "*_attr" params

* Fix: clear observations on reset

* Add velocity action manager and example (#26)

* observations.current_actions should return raw actions

* Add raw actions to wheeled robot example

* Fix action observation

* Fix: observation scale/noise was updating the original tensor value

* Add imu_projected_gravity

* Light cleanup

* Angular velocity indicator (#29)

* Add angular velocity indicators to the velocity command manager

* Update the wheeled robot to be a 4-wheel system

* Fix envs_idx to torch.Tensor type

* Fix action delay setup/reset

* Fix class name type: SkrlEnvWapper -> SkrlEnvWrapper
Misc typing fixes

* Automatically calculate sensitivity for velocity command rewards

* Add stopped_dof_velocity penalty

* Fix code error

* Update Freenove model and training environment

* Cleanup

* Update stopped_envs in velocity manager to take a threshold.

* More PR feedback/fixes

* Debug mesh builders (#32)

* Pose2d Command & Obstacles (#30)

* Update Freenove model and training environment

* Checkpoint

* Fix bugs

* Pose2dCommand

* Cleanup noise and needless constants.

* Cleanup readme, environment and unnecessary meshes

* Fix navigation training

* Remove unused rewards

* Fix: Observation noise is applied before scaling

* Update project settings

* Rename model directory

* Update/fix/run examples

* Update observation example

* Cleanup Pose2DCommand

* Cleanup

* Add pattern name matching utils

* copy edit

* lint trap

* Add pose2d videos
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant