Safety Gymnasium Versions Save

NeurIPS 2023: Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark

v1.2.0

8 months ago
  1. feat(safe isaac gym): upload heterogeneous FreightFranka robot as a new agent in PR #71
  2. feat(multi-agent): update multi-agent environments for the interaction of multiple agents in PR #70

v1.1.0

8 months ago
  1. Update building and formula_one tasks in PR #69
  2. We have incorporated a more realistic environment setting and provided a comprehensive exposition of the real-world relevance of the Goal, Push, and Button tasks, emphasizing their representativeness for practical applications in PR #69
  3. We introduced the FormulaOne racing environment. This environment is inspired by genuine F1 racing tracks and is characterized by intricate visual information, scenarios closely mirroring reality, and challenging tasks in PR #69

v1.0.0

9 months ago
  1. feat(environments): update safe issac gym environments in PR #65
  2. feat(vision env): upload safe vision environments in PR #61
  3. feat(masafe): merge multi-agent safe environments into main branch in PR #59

v0.4.1

10 months ago
  1. feat: support environments wrappers in PR #57

v0.4.0

1 year ago
  1. feat(agents): support Doggo agent in PR #51
  2. chore(benchmarks): upload benchmark results for environments in PR #48

v0.3.0

1 year ago
  1. test: enable tests in ci in PR #16
  2. chore: pre-commit autoupdate
  3. style: prefer utf-8 over UTF-8 in code
  4. feat: add Gymnasium conversion wrappers in PR #46
  5. chore: add test configurations and update Makefile in PR #45
  6. chore: update license header in PR #44
  7. style: fix grammar in README and normalize string in pyproject.toml
  8. deps(gymnasium): update gymnasium version to 0.28.1 in PR #41
  9. chore(pre-commit): [pre-commit.ci] autoupdate in PR #36
  10. chore: update link to OmniSafeAI org in PR #38

v0.2.0

1 year ago
  1. feat(SafeVelocity): update safety velocity tasks to v1 in PR #37

v0.1.3

1 year ago
  1. fix: fix the error that occurred in windows for default random seed in PR #32

v0.1.2

1 year ago
  1. feat: support rgb_array_list and depth_array_list mode for render in PR #26.
  2. fix: fix bug in vision observation in PR #28.
  3. chore(vision_env): update default resolution of vision environments in PR #29.

v0.1.1

1 year ago
  1. fix: fix origin AutoresetWrapper doesn't have cost (#23).
  2. fix: fix the bug in the observation space of the compass (#17).
  3. feat: enable more flake8 checks (#24)