Changelog#
Release v1.1.0 (2026-05-08)#
General#
The
_step_callbackof theBasicMagBotSingleAgentEnvnow returns the possibly modified action instead of None.Added support for Python 3.14
The
StateBasedGlobalPushingEnvand all related pushing environments, such asStateBasedPushBoxEnv,StateBasedPushTEnv,StateBasedPushLEnv, andStateBasedPushXEnvnow also support the rotation of the movers around the z-axis (\(\gamma\)-rotation) to establish more stable contacts with the object. It is possible, but not necessary, to specify a limit for the \(\gamma\)-rotation.
Performance#
Added spatial hashing for mover-mover collision detection.
Precompute per-tile wall-segment lookup table for wall collision checks.
Added fast paths for uniform mover sizes and complete tile layouts in
qpos_is_valid.Cached episode-level Shapely polygons in
StateBasedGlobalPushingEnvto avoid rebuilding them on every step.Removed
scipydependency and optimize rectangle vertex calculation ingeometry_2D_utils.
Bug Fixes#
Fixed wrong mover bumper mass and corresponding examples in the documentation
Fixed wrong gainprm in all RL environments
Ensured correct maximum dynamics of the actions (RL environments). This is fixed by using the
_step_callback.
Release v1.0.3 (2026-03-25)#
General#
Fixed bug in
StateBasedGlobalPushingEnv-v0Updated RL examples
Release v1.0.2 (2026-02-27)#
General#
Added the 6D-Platform MagBot (
SixDPlatformMagBotsAPM4330)Added tutorial on how to add MagBots to custom environments
Added example environments with MagBots (
SixDPlatformMagBotApplicationExampleEnvandSixDPlatformMagBotExampleEnv)
Release v1.0.1 (2025-11-24)#
General#
Used MuJoCo functions in
MoverImpedanceControllerinstead ofscipy.spatial.transform.Rotation(better performance, butscipy.spatial.transform.Rotationis not completely replaced)Added benchmark environments
Added
StateBasedPushXEnv-v0andStateBasedPushLEnv-v0Updated throughput calculation in pushing environments and update benchmark results accordingly
Release v1.0.0 (2025-11-08)#
General#
Added the basic environments:
BasicMagBotEnv,BasicMagBotSingleAgentEnv,BasicMagBotMultiAgentEnvAdded the following example environments for trajectory planning and object manipulation:
LongHorizonGlobalTrajectoryPlanningEnv-v0,StateBasedStaticObstaclePushingEnv-v0,StateBasedGlobalPushingEnv-v0,StateBasedPushBoxEnv-v0,StateBasedPushTEnv-v0Added a
MoverImpedanceControllerthat solves a position and orientation taskAdded the
MujocoOffScreenViewer: an extension of the Gymnasium OffScreenViewer that allows to also specify the groups of geoms to be rendered by the off-screen rendererAdded the
MujocoViewerCollection: a manager for all renderers in a MuJoCo environmentAdded the
Matplotlib2DViewer: a simple viewer that displays the tile and mover configuration together with the mover collision offsets for debugging and analyzing planning tasksAdded auxiliary functions for MuJoCo, collision checking and rotations
Added benchmarks and latest results for trajectory planning and object pushing
Added tutorials and examples on how to use MagBotSim