Skip to content

Releases: JeffA233/rlgym-sim-rs

v0.8.0-alpha

16 Jul 02:19
Compare
Choose a tag to compare

PRs/Merges

  • Changed math rand_vec3, it was generating all 0s always by @Kaiyotech in #5

Minor Breaking Change Notices

  • The most recent version of rocketsim_rs requires a new init argument
  • GameConfig now requires a car_config argument
  • GameMatch now has a use_single_obs argument
  • GameState now holds an array of rocketsim_rs::BoostPad for boost_pads instead of an array of floats
  • The RLViser packets have been updated to work with the latest version which will break pervious versions AFAIK

Notes

Full Changelog: v0.7.0-alpha...v0.8.0-alpha

Fixes:

  • AdvancedObs now has the correct inverted boost pads
  • Now clears previous_touched_ticks for every agent when a new config is introduced
  • PlayerData's on_ground is now updated to be true for tick_skip ticks instead of a singular tick no matter the tick_skip
  • Now fully resets the car and ball states to default when state setting on match resets
  • Multiple DefaultState and RandomState updates to prevent data from holding-over from the previous match

Minor changes:

  • Bumped rocketsim_rs version

New features:

  • State setting should now be possible within RLViser when spectating an environment
  • Pausing and match speed adjustments can also be done via RLViser although match speed appears to carry over from the previous instance upon starting potentially
  • Added serde's serialization and deserialization macros to a number of informational/data-holding structs
  • The car body in RocketSim can now be changed via the car_config argument in GameConfig, it was previously always Octane

v0.7.0-alpha

20 Dec 04:01
Compare
Choose a tag to compare
v0.7.0-alpha Pre-release
Pre-release

PRs/Merges

  • change get_action_space to usize to match obs_space by @Kaiyotech in #4

Depreciation Notice

  • Previous action was removed from the reward functions and observation builders in favor of the supplied PlayerData last_actions

Breaking Change Notice

  • GameMatch's get_config() return type has been changed from an array of values to just simply a new GameConfig

Notes

Full Changelog: v0.6.0-alpha...v0.7.0-alpha

Fixes:

  • Updated the way the wrapper changes the arena to match new game configs which should fix any oddities with team size changes
  • Fixed how agent assists are checked for validity
  • The renderer now works with tick skip instead of skipping states and having an incorrect update rate with a tick skip other than 1

Minor changes:

  • Bumped rocketsim_rs version

New features:

  • Added From implementation for Velocity and Position for easy conversion if necessary
  • Added more docs

v0.6.0-alpha

01 Dec 05:32
Compare
Choose a tag to compare
v0.6.0-alpha Pre-release
Pre-release

PRs/Merges

  • add replay setter and tests, modify gitignore and cargo to support it by @Kaiyotech in #1
  • adding into_array trait to carcontrols by @Kaiyotech in #2
  • Render branch merge by @JeffA233 in #3

New Contributor

Depreciation Notice

  • The previous action input for reward functions and observation builders will be removed in favor of the controls being in the PlayerData in the next version.

Notes

Full Changelog: v0.5.0-alpha...v0.6.0-alpha

Fixes:

  • Updated random state's X/Y max to not clip into walls potentially
  • Fixed rotation matrix bug
  • Car IDs are now properly matched in the state setter to the corresponding RocketSim car

Minor changes:

  • Added debug to multiple structs
  • Made ID and team num in car wrapper immutable
  • Observation builder assert now just checks for at least enough builders and not exactly

New features:

  • @Kaiyotech added a replay setter that can use Numpy files
  • Added rendering support with RLViser
    • Note that this will break things as make now requires a new RenderConfig struct!
  • @Kaiyotech Added into_array() for CarControls
  • Added f32 ops to Position and Velocity

v0.5.0-alpha

27 Nov 06:44
Compare
Choose a tag to compare
v0.5.0-alpha Pre-release
Pre-release

Full Changelog: v0.4.0-alpha...0.5.0-alpha

  • Added previous actions to player struct, will remove previous action from reward function and observation builder likely soon because of this
  • Fixed tester.rs import
  • Fixed bug regarding previous action not updating to fit a new number of agents
  • Mapped car ids to always be from 1 to 6 in the gym instead of getting them directly from RocketSim which kept ticking up forever when removing and adding new cars

v0.4.0-alpha

10 Nov 07:53
Compare
Choose a tag to compare
v0.4.0-alpha Pre-release
Pre-release

Full Changelog: v0.3.0-alpha...v0.4.0-alpha

  • Added more docs
  • Added WeightedSampleSetter
  • Added CombinedTerminalConditions and NoTouchKickoffTimeoutCondition (long names, may change in future)
  • Switched from StdRng to SmallRng for certain random operations

v0.3.0-alpha

09 Nov 01:27
Compare
Choose a tag to compare
v0.3.0-alpha Pre-release
Pre-release

Full Changelog: v0.2.0-alpha...v0.3.0-alpha
-bumped RocketSim version to fix incorrect scoring problem
-fix and update math funcs
-update state generator
-boost pad position fix

v0.2.0-alpha

20 Oct 20:01
Compare
Choose a tag to compare
v0.2.0-alpha Pre-release
Pre-release

Full Changelog: v0.1.0-alpha...v0.2.0-alpha
-fixed rotation matrix inversion
-added state generator prototype
-added boost pads to state wrapper

v0.1.0-alpha

20 Oct 05:18
Compare
Choose a tag to compare
v0.1.0-alpha Pre-release
Pre-release

Everything needed to get started should now be in this version and hopefully with few if any bugs. It is mostly undocumented for now, so there is not an easy way to learn about it other than knowledge from the RLGym side in Python and looking through the APIs/examples. Documentation should come soon in future versions.