Beta Announcement and Roadmap to 1.0

[momo-van]

Beta 2 Release Highlights

This Beta showcases significant progress across Newton’s solvers, collision, sensing, and control modules, and demonstrates how the core building blocks are coming together through new examples. The highlights below capture the most notable capabilities added in this release.

1. Collision pipeline and hydroelastic contacts: A flexible collision pipeline optimized for scenes with dynamic topology or large numbers of shapes, supporting multiple broadphase strategies. The pipeline also introduces SDF-based collisions, and Drake-style hydroelastic contacts enabling robotic manipulation and precision-assembly tasks, as well as non-uniform terrain modeling using triangular meshes.
2. Tiled camera sensor: Introduced batched/tiled camera rendering supporting high-throughput low-fidelity perception workflows, alongside expanded sensor module support including particle-aware raycasting and site sensors.
3. Maturing deformable simulation, introducing cables: Expanded VBD solver beyond surface/thin deformable by adding linear deformable capability for cable modeling and simulation, also demonstrated two-way coupling between XPBD rigid bodies and MPM materials.
4. Enhanced MuJoCo Warp integration: Improved solver defaults, stability, and benchmarking, with added flexibility to run MuJoCo Warp using Newton’s unified collision pipeline with SDF-based collision handling, and hydroelastic contacts.
5. Stronger IK and optimization workflows: Upgrades the IK solver with L-BFGS, improving convergence and making Newton a more effective building block for optimization-driven control and design.
6. Improved visualization and inspection: More predictable viewer behavior (offset handling, inertia visualization, mesh logging) and continued Rerun backend refinements improve debugging, analysis, and iteration workflows.
7. Expanded getting-started examples: New examples accompany major Beta features, making it easier to adopt tiled sensing, multiphysics coupling, contact, and deformable workflows.

Roadmap to 1.0

We’ve laid the groundwork in Newton with foundational modules, including solvers, sensors, controls, collision, among others. While this release has shown what’s possible with these components (like solver coupling and new examples) our focus for 1.0 is on delivering fully integrated, end-to-end workflows. In particular, we’re focusing on high-fidelity simulations of dexterous robotic manipulation for industrial tasks. We’re moving from simple examples to real-world scenarios that ensure the readiness of these workflows, while providing a first-class developer experience, stable API and effective documentation.

Feature highlight for 1.0 release:

1. Newton USD Schema: Introduce a Newton USD schema for both rigid and deformable bodies, serving as a common data interchange for authoring and simulation workflows.
2. Finalized Public API: Complete and stabilize Newton’s API.
3. Complete MuJoCo Warp Integration: Finish the MuJoCo Warp integration and optimization to ensure that parsing from USD, MJCF, and URDF mirrors native MuJoCo behavior.
4. Kamino Solver: Release the Kamino rigid-body maximal coordinate solver to support complex mechanisms, including closed kinematic loops.
5. Deformables: Advance cable support in the VBD solver, also extend VBD to support 3D elastic solids, completing Newton’s deformable spectrum from 1D and 2D to full volumetric models.
6. Deformable-Rigid Coupling: Provide first-class two-way coupling between deformable and rigid body solvers, particularly VBD and MuJoCo Warp, with special focus on manipulation scenarios.
7. Actuator Library: Add an actuator interface and a curated actuator library within the Newton ecosystem for consistent, modular integration of actuation into simulation.
8. Isaac Lab & Isaac Sim Integrations: Release Newton as the default physics backend for Isaac Lab 3.0 and provide a developer preview integration with Isaac Sim 6.0.

As we work towards 1.0, we’re committed to making Newton the backbone for robotics simulation in reinforcement learning, imitation learning, synthetic data generation, and software-in-the-loop testing, both as a standalone library and as part of integrated Isaac workflows.

Newton will continue to evolve and mature over time, extending well beyond the 1.0 release. We’d love to hear your feedback on both the Beta release and the roadmap, please keep it coming and help shape Newton’s future.

Ecosystem Adoption

The Newton open ecosystem continues to grow, with universities and companies adopting Newton to integrate specialized solvers and workflows. These efforts span areas such as tactile sensing, cloth simulation, dexterous manipulation, and rough-terrain locomotion, demonstrating how Newton serves as a shared foundation for robotic learning and helps bridge the sim-to-real gap.

Learn more about early adopters, Newton’s architecture, highlights from an earlier Beta, and practical examples in this technical blog.

[rdsa-nv]

We also want to share a recent announcement regarding Newton and Isaac Lab integration found here: [Important] Isaac Lab Development Update
See additional details related to Newton under: Feature Branches > feature/newton corresponding to the Isaac Lab - Newton Beta 2 branch.