Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

[0.5.0] - 2026-08-09

Added

  • A responsive DHB-XR identity system with separate display, compact, navigation, and 24 px masters for light and dark documentation surfaces. The production mark keeps the frame brackets, reusable motion path, and wordmark separate so the identity remains clear at documentation and favicon sizes.
  • A package-style overview plus a minimal frame-invariance example whose technical figure shows changed world coordinates, overlapping initial-frame geometry, and DHB residuals below a declared numerical tolerance.
  • VLA guidance that positions initial-frame relative actions as the strong baseline, then distinguishes DHB-XR through reusable motion geometry, retargeting, retrieval, time normalization, and action-token interfaces.
  • Merge-triggered, trusted PyPI publishing with synchronized release-metadata validation, immutable tag checks, separately retained build artifacts, and a GitHub release created only after publication succeeds.

Development preview

The following additions have software or isolated component coverage but are not validated simulator or robot capabilities. They are documented separately from the release highlights.

  • Object-centric, source-traceable skill capsules that retain raw world motion, TCP-in-object motion, sparse relations, end-effector actions, source dataset, revision, license, and reproducible multiscale DHB descriptor indexes.
  • Typed intermediate relational-pose anchors, a common anchor gate, persistent matched FATROP regularizers, and a strong piecewise raw-SE(3) anchor control for representation-isolation experiments.
  • A strong endpoint-corrected raw-SE(3) reuse control, an orientation-aware payload/fixture clearance gate, and an exploratory matched retarget-isolation benchmark for identifying where DHB-XR geometry is actually distinctive.
  • Typed manipulation primitive contracts spanning SE(3) motion, portable end-effector commands, observation checks, evidence, and program graphs with explicit recovery paths.
  • Minimum-jerk, affordance-normal, arc, screw, and shared-waypoint analytic paths.
  • Deterministic primitive-program execution with bounded retries, reobservation, stale-epoch cancellation, and concise typed commands.
  • Evidence-preserving primitive memory, sample-exact and hysteretic gripper segmentation, shared transition-waypoint utilities, and adapters for DHB/OCP and learned candidate sources.
  • Canonical demonstration-trace ingestion that preserves dataset, license, action convention, raw-signal source details, and requires explicit semantics before candidate promotion.
  • A matched primitive portfolio with common epoch, endpoint, and sampled-TCP clearance gates plus a reproducible 20-trial component benchmark.
  • A runnable architecture example and primitive-program/API documentation.
  • A release-value and evidence matrix, complete primitive public API inventory, and JSON-derived PNG/SVG dashboards for isolated primitive benchmarks.

Changed

  • CuSADi batch decoding now keeps the translational DHB frame independent from tool orientation and accepts explicit per-trajectory initial linear frames.

Development-preview changes

  • Persistent relational FATROP now transcribes a reusable CasADi solve function once, so steady-state requests avoid repeated Rockit value, initialization, and solve-surface setup. The dense oriented-payload SAT gate is vectorized across path samples and fixtures with scalar parity coverage.
  • Package and primitive exports are lazy, so contract/executor integrations do not import SciPy, plotting, or native solver dependencies at startup.
  • Backend results that explicitly report success=False now fail closed instead of being promoted into gateable primitive candidates.

Fixed

  • Restored missing type imports in the Torch decoder and DHB-TI annotations, and made hierarchical residual initialization explicit.

Development-preview fixes

  • During-motion end-effector actions now honor both ends of their inclusive phase window and fail closed with a motion-stop command if feedback skips past an unissued action.
  • Isolated retarget benchmark reports now serialize NumPy-valued nested gate metrics instead of failing after all trials have completed.

[0.4.1] - 2026-05-27

Added

  • Fixed-horizon CusADi decode artifacts and generated CUDA source for 50, 80, 100, 150, and 200 sample trajectories.
  • dhb_xr-build-cusadi-decode for compiling supported CusADi decode libraries into a dhb_xr-owned cache.
  • GPU decode documentation covering install paths, cache behavior, runtime fallback policy, and deployment troubleshooting.

Changed

  • generate_trajectory(..., backend="auto") documentation now presents FATROP as the preferred single-trajectory backend with explicit fallback policy.
  • The cusadi optional extra now depends on CasADi and PyTorch rather than an external editable CusADi checkout.

[0.4.0] - 2026-05-14

Added

  • Torch DHB-QR decoder path.
  • VLA action-head, chunk rollout, SmolVLA adapter, and waypoint controller utilities.
  • Focused tests for action-head behavior, chunk rollout continuity, VLA pipeline stages, LIBERO orientation handling, Torch DHB-QR decoding, OAT tokenizer diagnostics, waypoint control, and solver retargeting.

Changed

  • Improved VLA pipeline token streaming and OAT prefix decoding behavior.
  • Updated CasADi/Fatrop retargeting paths and timing documentation.
  • Standardized VLA dataset adapter outputs on DHB-XR's public wxyz quaternion convention.
  • Removed unused dtaidistance from database extras and included einops in the all extra.

Fixed

  • LIBERO quaternion/action semantics and rollout orientation handling.
  • Fatrop boundary handling and trajectory-derived initial frame usage in retargeting.

[0.3.0] - 2026-01-30

Added

  • FAST Tokenizer: Frequency-space action tokenizer using DCT + BPE compression, no training required (~28x compression on invariant sequences)
  • OAT-style Tokenizer: Ordered Action Tokenization with Register Encoder, Finite Scalar Quantization (FSQ), and Masked Nested Dropout for anytime prefix decoding
  • FSQ Module: Finite Scalar Quantization with deterministic rounding -- eliminates VQ codebook collapse by construction
  • Register Encoder: Transformer-based temporal compression using learnable register tokens with causal-last attention
  • Nested Dropout: Masked nested dropout module for inducing coarse-to-fine token ordering (Matryoshka-style)
  • OAT Decoder: Cross-attention Transformer decoder for reconstructing sequences from register token latents
  • Benchmark Suite: Three benchmark scripts comparing all 7 tokenization methods on rate-distortion, SE(3) stability, and prefix decoding
  • Interactive Notebook: Comprehensive tutorial notebook (tutorial_action_tokenization.ipynb) with head-to-head comparisons and visualizations
  • Manuscript Updates: Expanded Section 7 with FAST/OAT descriptions, new Appendix with empirical comparison tables

Fixed

  • SE(3) Invariance: Fixed invariant token stability under SE(3) transforms by using trajectory-relative initial frames (init_pose from trajectory start) instead of fixed global axes. This achieves exact invariance (0% token change, MSE = O(1e-13)) versus the previous approximate invariance
  • BPE Token Comparison: Improved token stability reporting to account for different-length BPE sequences and separate pre-BPE (DCT integer) from post-BPE token differences

Changed

  • einops dependency: Added to [tokenization] optional dependencies for FSQ implementation

[0.2.0] - 2026-02-05

Added

  • VLA Integration: Comprehensive LIBERO and RoboCASA support for Vision-Language-Action benchmarks
  • LIBERO-PRO Support: Full integration with LIBERO-PRO perturbation robustness testing
  • Trajectory Adaptation: Fatrop and CasADi solvers for trajectory retargeting (decode ~5-10ms, Fatrop NLP <1s)
  • Swap Demo: Compelling demonstration showing DHB-XR vs naive replay under spatial perturbations (6.5cm improvement)
  • Simulation Integration: Real-time LIBERO environment execution with OpenCV visualization and video recording
  • VLA Documentation: Extensive documentation explaining how DHB-XR addresses VLA spatial generalization challenges

Changed

  • Repository URLs: Updated all links to point to robodreamer organization
  • Documentation Updates: Added VLA context throughout README and documentation

Fixed

  • Solver Integration: Proper Fatrop/CasADi solver selection in trajectory adaptation functions

[0.1.2] - 2026-02-04

Added

  • MkDocs Documentation: Migrated from Sphinx to MkDocs with Material theme
  • GitHub Actions CI/CD: Automated documentation builds and deployment
  • Encoding Method Enum: Replaced string literals with type-safe EncodingMethod enum
  • GitHub Pages: Documentation hosting setup

Changed

  • Project Metadata: Updated author email and added scikit-learn dependency
  • Copyright Year: Updated to 2026
  • Pixi Configuration: Renamed project workspace

[0.1.1] - 2026-02-03

Added

  • Trajectory Preprocessing: Robustness modules for handling reversals and zero-motion segments
  • DHB-Token: VQ-VAE tokenization pipeline for discrete action representations
  • Generative Models: Variational Flow Matching (VFM) for multi-modal trajectory generation
  • Motion Database: Similarity search and retrieval with DTW support

[0.1.0] - 2026-01-31

Added

  • Core DHB-XR Implementation: DHB-DR (Double-Reflection) and DHB-QR (Quaternion-Relative) invariant encoding
  • Time-Invariant Reparameterization: DHB-TI for speed-independent trajectory representations
  • GPU Acceleration: PyTorch-based batched operations and CusADi for large-scale optimization
  • Fatrop Integration: Fast structured OCP solver for trajectory optimization
  • CasADi Support: General nonlinear optimization for trajectory adaptation
  • Visualization: Enhanced SE(3) trajectory plotting with matplotlib
  • Jupyter Notebooks: Comprehensive demo notebook with interactive examples

Changed

  • Dependencies: Added CUDA support with PyTorch 2.0+ and optional GPU features
  • Documentation: Enhanced README with performance benchmarks and usage examples

[0.0.1] - 2026-01-30

Added

  • Initial Project Structure: Basic DHB encoding/decoding functionality
  • Project Configuration: Pixi environment, GitHub Actions, and development tooling
  • Basic Testing: Initial test suite setup
  • Documentation: Sphinx-based documentation framework