Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.

This product includes software developed by third parties:

--------------------------------------------------------------------------------
strands_robots/policies/cosmos3/_msgpack_numpy.py

Adapted from openpi-client (https://github.com/Physical-Intelligence/openpi)
Original code by Physical Intelligence and contributors.
Licensed under the Apache License, Version 2.0.

The code is itself adapted from msgpack-numpy
(https://github.com/lebedov/msgpack-numpy) by Lev E. Givon.
Licensed under the BSD License.

You may obtain a copy of the Apache License at:
    http://www.apache.org/licenses/LICENSE-2.0
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
Optional Cosmos 3 -> MuJoCo sim-loop bridge (strands_robots/policies/cosmos3/sim_ik.py,
strands_robots/policies/cosmos3/action_decode.py)

The optional Cosmos3Policy(backend="diffusers") sim-loop bridge turns the model's
raw unified action into MuJoCo joint targets (de-normalize -> EE-pose decode ->
inverse kinematics) using:

  * mink (https://github.com/kevinzakka/mink)
    Apache License, Version 2.0. Differential inverse kinematics on the MuJoCo
    model (FrameTask + PostureTask).
  * MuJoCo (https://github.com/google-deepmind/mujoco)
    Apache License, Version 2.0.

The per-embodiment action-normalization quantiles bundled under
strands_robots/policies/cosmos3/stats/ (q01/q99) are derived from NVIDIA's
cosmos_framework (data/vfm/action/datasets/stats); the de-normalization formula
mirrors cosmos_framework's denormalize_action(method="quantile"). NVIDIA Cosmos
is distributed by NVIDIA under its respective licenses; consult the model card.
These are optional runtime dependencies and are not redistributed with this
package.

You may obtain a copy of the Apache License at:
    http://www.apache.org/licenses/LICENSE-2.0
--------------------------------------------------------------------------------

--------------------------------------------------------------------------------
Optional in-process Cosmos 3 backend (strands_robots/policies/cosmos3/policy_diffusers.py)

The optional Cosmos3Policy(backend="diffusers") path loads NVIDIA Cosmos 3
world-foundation models in-process via:

  * Hugging Face diffusers (https://github.com/huggingface/diffusers)
    Apache License, Version 2.0. The Cosmos3OmniPipeline + CosmosActionCondition
    used here ship in diffusers-from-source.

NVIDIA Cosmos models (e.g. nvidia/Cosmos3-Nano) are distributed by NVIDIA under
their respective model licenses; consult the model card for terms. These are
optional runtime dependencies and are not redistributed with this package.

You may obtain a copy of the Apache License at:
    http://www.apache.org/licenses/LICENSE-2.0
--------------------------------------------------------------------------------

--------------------------------------------------------------------------------
Optional WBC locomotion policy (strands_robots/policies/wbc/)

The optional WBCPolicy ([wbc] extra) is a clean-room implementation of the
control loop described by NVIDIA's GR00T Whole-Body-Control reference:

  * GR00T-WholeBodyControl (https://github.com/NVlabs/GR00T-WholeBodyControl)
    Apache License, Version 2.0. The observation layout, PD-control law, and
    SONIC controller contract reproduced in strands_robots/policies/wbc/ are
    derived from the upstream reference runner
    (decoupled_wbc/sim2mujoco/scripts/run_mujoco_gear_wbc.py). No upstream code
    is vendored.

  * ONNX Runtime (https://github.com/microsoft/onnxruntime)
    MIT License. Runs the controller's ONNX sessions in-process.

NVIDIA GR00T-WBC model weights (e.g. nvidia/GEAR-SONIC) are distributed by
NVIDIA under the NVIDIA Open Model License; consult the model card for terms.
These are optional runtime dependencies and are not redistributed with this
package.

You may obtain a copy of the Apache License at:
    http://www.apache.org/licenses/LICENSE-2.0
--------------------------------------------------------------------------------

--------------------------------------------------------------------------------
From-scratch RL trainers (strands_robots/training/rl/base_algo.py,
strands_robots/training/rl/normalization.py,
strands_robots/training/rl/ppo.py,
strands_robots/training/rl/fast_sac.py,
strands_robots/training/rl/replay_buffer.py)

The from-scratch reinforcement-learning trainers reproduce control-and-statistics
math from:

  * Amazon FAR Holosoma (https://github.com/amazon-far/holosoma)
    Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
    BSD-3-Clause License. The PPO algorithm (clipped policy surrogate, clipped
    value loss, Generalized Advantage Estimation, advantage normalization), the
    FastSAC algorithm (clipped double-Q target, tanh-squashed Gaussian actor,
    automatic entropy temperature, SimpleReplayBuffer), the BaseAlgo lifecycle,
    and the EmpiricalNormalization (Welford) observation normalizer are adapted
    from this project, re-homed onto the strands-robots SimEngine env interface.
    Holosoma is itself derived in part from RSL-RL
    (https://github.com/leggedrobotics/rsl_rl), BSD-3-Clause.

  The BSD-3-Clause License text is reproduced below:

    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions are met:

    1. Redistributions of source code must retain the above copyright notice,
       this list of conditions and the following disclaimer.
    2. Redistributions in binary form must reproduce the above copyright notice,
       this list of conditions and the following disclaimer in the documentation
       and/or other materials provided with the distribution.
    3. Neither the name of the copyright holder nor the names of its
       contributors may be used to endorse or promote products derived from this
       software without specific prior written permission.

    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
    AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
    IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
    ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
    LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
    CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
    SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
    INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
    CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
    ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
    POSSIBILITY OF SUCH DAMAGE.
--------------------------------------------------------------------------------
