pith. machine review for the scientific record. sign in

arxiv: 2603.05493 · v2 · submitted 2026-03-05 · 💻 cs.RO

Recognition: 2 theorem links

· Lean Theorem

cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots

Balakumar Sundaralingam , Adithyavairavan Murali , Stan Birchfield
Authors on Pith no claims yet

Pith reviewed 2026-05-15 15:41 UTC · model grok-4.3

classification 💻 cs.RO
keywords motion planningGPU accelerationsigned distance fieldshumanoid robotstrajectory optimizationinverse kinematicswhole-body controlB-splines
0
0 comments X

The pith

cuRoboV2 unifies B-spline optimization, dense GPU distance fields, and scalable whole-body computations to generate reliable motions for high-DoF robots.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces cuRoboV2 as a single framework that ties together trajectory planning, environment perception, and whole-body robot control. It uses B-spline curves to produce smooth paths that obey torque limits, a fast GPU pipeline to build dense signed distance fields from depth data for collision checking, and efficient kinematics and dynamics calculations that extend to humanoid systems. These pieces together deliver much higher task success rates and speed than prior separate tools. A reader would care because current robot software often breaks when moving from simple arms to complex bodies or from simulation to real scenes.

Core claim

cuRoboV2 achieves 99.7 percent success under 3kg payload, 99.6 percent collision-free inverse kinematics on a 48-DoF humanoid, and 89.5 percent retargeting constraint satisfaction through B-spline trajectory optimization that enforces smoothness and torque limits, a GPU-native TSDF/ESDF pipeline that produces dense signed distance fields up to 10 times faster and with 8 times less memory than prior work, and scalable whole-body GPU computations including topology-aware kinematics, differentiable inverse dynamics, and map-reduce self-collision detection that yield up to 61 times speedup.

What carries the argument

The GPU-native TSDF/ESDF pipeline that generates dense signed distance fields covering the full workspace, paired with map-reduce self-collision detection and differentiable inverse dynamics for whole-body control.

Load-bearing premise

The perception pipeline and whole-body computations remain accurate and real-time when transferred from simulation or controlled benchmarks to unstructured real-world environments with sensor noise, calibration errors, and dynamic obstacles.

What would settle it

A physical test on a 48-DoF humanoid in a room with moving obstacles and noisy depth sensors where collision-free motion success falls below 80 percent would show the scaling claim does not hold.

Figures

Figures reproduced from arXiv: 2603.05493 by Adithyavairavan Murali, Balakumar Sundaralingam, Stan Birchfield.

Figure 1
Figure 1. Figure 1: Trajectory optimization loop. Each iteration evaluates B-spline waypoints, computes kinemat￾ics and inverse dynamics in parallel, evaluates costs (scene collision, self-collision, configuration-space bounds) concurrently on separate CUDA streams, aggregates them, backpropagates gradients through all forward operations, and updates the B-spline control points via L-BFGS [PITH_FULL_IMAGE:figures/full_fig_p0… view at source ↗
Figure 2
Figure 2. Figure 2: Local support property of cubic B-splines. Perturbing knot 𝑢3 only affects 4 neighboring curve segments (orange region). Interpolation points (dots) are uniformly spaced in time; during optimization, each point within this region contributes a gradient to 𝑢3 (accumulated with GPU warp-level reductions). (For simplicity, 𝑁interp = 2 is shown.) Given the control points 𝑢𝑘, 𝑘 = 0, . . . , 𝐾 − 1 for a particul… view at source ↗
Figure 3
Figure 3. Figure 3: TSDF-ESDF pipeline. Depth images are fused into a block-sparse TSDF via voxel-centric projec￾tion (Sec. 5.2), and known geometry is stamped analytically. After each update, TSDF weights are decayed (time + frustum factors), and blocks below a weight threshold are tombstoned and recycled. On demand, a dense ESDF is generated in three stages: (1) seeding surface sites from zero-crossings, (2) propagating dis… view at source ↗
Figure 4
Figure 4. Figure 4: Block-sparse TSDF storage. Only blocks near observed surfaces are allocated. A hash table maps block coordinates to pool indices via CAS. Each voxel stores two independent float16 channels (depth and geometry) whose minimum is returned at query time. Recycled blocks are managed through a free-list stack. 5.2. Depth and Primitive Integration (a) Block Discovery Depth Blocks per-pixel rays 𝑁 =px×samp threads… view at source ↗
Figure 5
Figure 5. Figure 5: Voxel-Project depth integration. (a) Per-pixel rays discover blocks touched by the current depth frame. (b) Duplicate keys are filtered and surviving blocks are allocated via CAS, yielding 𝐾 pool indices. (c) Phase 4 reverses the mapping: each voxel projects itself into the image, reads the depth at the projected pixel, and writes the signed distance directly, eliminating all atomic contention. Depth integ… view at source ↗
Figure 6
Figure 6. Figure 6: ESDF generation pipeline. (a) Site seeding: scatter launches one thread per TSDF voxel and writes to the ESDF (requires atomics); gather launches one thread per ESDF voxel and probes the TSDF via a 7-point stencil (no atomics, CUDA graph safe). (b) PBA+ propagates site ownership via separable sweeps: Phase 1 sweeps rows via bidirectional flooding, Phase 2 builds Maurer stacks along columns to produce the e… view at source ↗
Figure 7
Figure 7. Figure 7: Kinematic structures that challenge parallelized computation. (a) A single-arm serial chain, the traditional case. (b) Mimic joints, where actuating 𝜃1 mechanically constrains 𝜃2 and 𝜃3. (c) A humanoid with branching kinematic chains. We introduce precomputed topology caches for 𝑂(1) ancestor lookups and two-stage Jacobian filtering to efficiently parallelize gradient and Jacobian computation. 6.1. Kinemat… view at source ↗
Figure 8
Figure 8. Figure 8: Adaptive forward kinematics execution. For simple robots, a single fused kernel computes frame transforms, sphere positions, and Jacobians. For complex robots, this is split into two kernels: the first computes frame transforms, and the second computes sphere positions and Jacobians. 6.1.2. Parallel Gradient Backpropagation The adaptive kernel dispatch described above accelerates forward kinematics, but ba… view at source ↗
Figure 9
Figure 9. Figure 9: Two-stage self-collision checking. For complex robots, collision pairs are partitioned across GPU blocks (map). A first kernel finds local maxima, and a second kernel reduces these to find the global maximum. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: RNEA forward and VJP backward kernels. We use Featherstone’s spatial algebra [15]: 𝑋𝑘 is the spatial transform from parent to child frame, 𝐼𝑘 the spatial inertia, and 𝑆𝑘 the joint motion subspace (a unit 6-vector for 1-DoF joints). The operators crm(·) and crf(·) denote the 6×6 motion and force cross-product matrices, with shorthand 𝑣 ×𝑚 𝑢 = crm(𝑣) · 𝑢 and 𝑣 × ∗ 𝑓 = crf(𝑣) · 𝑓 . Bar notation (𝑣¯, ¯𝑓 ) ind… view at source ↗
Figure 11
Figure 11. Figure 11: Motion Planning Results (75th percentile). Top: Kinematic success (solid) vs. dynamics success with 3 kg payload (hatched). Methods without dynamics constraints fail when torque limits are checked, even at zero payload. Middle/Bottom: Our cuRoboV2 achieves the highest payload success (99.7%) and lowest energy (106 J), without sacrificing other trajectory quality metrics. We evaluate our B-spline trajector… view at source ↗
Figure 12
Figure 12. Figure 12: Trajectory profiles for the same motion planning problem (each color denotes one joint). Left: cuRobo’s per-timestep optimization produces discontinuous acceleration and jerk. Right: cuRoboV2’s B￾spline optimization yields smooth, continuous derivatives suitable for real robot execution [PITH_FULL_IMAGE:figures/full_fig_p021_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Inverse dynamics performance (forward + backward pass) across batch sizes. cuRoboV2 is 1.5– 2× slower than GRiD’s code-generated kernels on simple robots but 14–18× faster than Newton. Critically, GRiD fails on the 48-DoF humanoid due to shared memory limits, while cuRoboV2 scales gracefully. 7.4. High-DoF Inverse Kinematics We evaluate the scalability of our kinematics and self-collision (Sec. 6) across … view at source ↗
Figure 15
Figure 15. Figure 15: IK results (batch size 100). (a) Standard IK: all methods achieve near-perfect success on single￾arm robots; on the 48-DoF humanoid, cuRoboV2 reaches 100% success while PyRoki drops to 49.8%. (b) Self￾collision-free IK: cuRoboV2 achieves 99.6% success on the humanoid via LM seeding followed by L-BFGS refinement; cuRobo and PyRoki both fail completely (0%). 24 [PITH_FULL_IMAGE:figures/full_fig_p024_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Retargeting visual comparison across GMR (top), PyRoki (middle), and cuRoboV2 (bottom) during a crawling sequence. The sequence requires hands to cross: GMR produces self-collisions (see red circle) since collision avoidance is not activated, PyRoki avoids collision but fails near contact boundaries, preventing the hands from crossing. cuRoboV2 resolves self-collisions gracefully, achieving accurate non￾c… view at source ↗
Figure 17
Figure 17. Figure 17: Humanoid motion retargeting on the G1 (70k frames). cuRoboV2-IK achieves 89.5% constraint satisfaction vs. PyRoki 61.2%, mink 54.5%, and GMR 40.6%, thanks to LM+L-BFGS seeding for collision-free solutions. MPC reaches 96.6% by checking collisions between frames but at the cost of tracking accuracy. compared to 65% for cuRoboV2 and 70% for GMR. These short episodes starve the policy of learning signal: the… view at source ↗
Figure 18
Figure 18. Figure 18: (a) Policy training curves (normalized to %). On running (top), PyRoki plateaus at 31% episode length due to repeated falls. On crawling (bottom), cuRoboV2, GMR, and mink all appear similar, masking the differences revealed by evaluation. (b) Evaluation box plots across 5 seeds. cuRoboV2 achieves the lowest MPJPE with tight variance, while GMR exhibits 12× higher MPJPE variance on crawling. 28 [PITH_FULL… view at source ↗
Figure 19
Figure 19. Figure 19: Learned locomotion policies on the Unitree G1 (green: cuRoboV2, pink: mink, orange: PyRoki, purple: GMR). (a) Running: PyRoki falls at high velocity; the other three methods maintain stable gaits. (b, c) Crawling across two sequences: GMR exhibits high seed-to-seed variability, tumbling in (b) and collapsing in (c), consistent with its 13× higher MPJPE variance. PyRoki and mink perform better with collisi… view at source ↗
Figure 20
Figure 20. Figure 20: ESDF benchmark across TSDF resolutions (10–25 mm) at 50% and 100% workspace cover￾age. cuRoboV2 achieves 2–10× lower total time and 2–8× less memory than nvblox across all configu￾rations, while maintaining higher collision recall. TSDF integration time is comparable across methods; the speedup stems from PBA+ distance propagation. Recall is computed only over the region where nvblox produces valid distan… view at source ↗
Figure 21
Figure 21. Figure 21 [PITH_FULL_IMAGE:figures/full_fig_p031_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Signed distance field cross-section (4 m × 4 m slice) at 100% workspace coverage. nvblox (mid￾dle) produces distances only within allocated blocks, leaving gaps in unobserved regions. cuRoboV2 (right) covers the full workspace, closely matching the ground-truth SDF (left). Unlike nvblox, which requires matching TSDF and ESDF resolutions, our method supports heteroge￾neous resolutions: we maintain a fine T… view at source ↗
Figure 23
Figure 23. Figure 23: Real-world deployment pipeline. Depth from a ZED Mini stereo camera is segmented to remove the robot, fused into a TSDF, and converted to an ESDF for real-time collision avoidance in the MPC trajectory optimizer [PITH_FULL_IMAGE:figures/full_fig_p033_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Real-world collision avoidance on the I2RT YAM robot. The robot plans a motion from left to right while cuRoboV2’s MPC dynamically avoids an obstacle using live ESDF updates from a ZED Mini stereo camera. 8. LLM-assisted Development We hypothesize that a well-structured codebase is a prerequisite for productive LLM-assisted development, and that the investment pays for itself. By restructuring cuRoboV2 fo… view at source ↗
Figure 25
Figure 25. Figure 25: Codebase size comparison (excluding blank lines). Left: Non-blank lines split into Code, Inline￾docs, and Tests, stacked by language. Right: File counts by language [PITH_FULL_IMAGE:figures/full_fig_p034_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Percentage of LLM-authored versus human-authored code additions per development phase. Early phases (R1–R3) focused on refactoring, with the human writing most code. Later phases (N1–N3) developed new modules using LLM-assisted workflows, with LLM contributions rising to 73% [PITH_FULL_IMAGE:figures/full_fig_p035_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Human–LLM workflow for RNEA development. The human (top) provides the NumPy for￾ward reference and suggests kernel optimizations; the LLM (bottom) handles each subsequent stage from derivation through profiling. (∼800 lines across forward and backward kernels, plus ∼300 lines of spatial algebra helpers) and wired the module into the CUDA Python backend with a PyTorch autograd wrapper (∼330 lines). Kernel … view at source ↗
Figure 28
Figure 28. Figure 28: Unified scene collision architecture. Left: cuRoboV1 splits collision across two kernel backends and a deep inheritance chain. Right: cuRoboV2 unifies all obstacle types under a single type-generic Warp kernel, reducing total code by ∼50%. Scene Collision Checking Migration to Warp. cuRoboV1 split collision checking across two backends: a monolithic CUDA kernel (∼2,600 code lines) for cuboids and voxels, … view at source ↗
Figure 29
Figure 29. Figure 29: Human–LLM workflow for migrating scene collision checking to Warp. Orange boxes (■) denote human contributions; blue boxes (■) denote LLM contributions. The migration proceeded in four phases ( [PITH_FULL_IMAGE:figures/full_fig_p037_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: Installation workflow comparison. Top: v1’s pybind11 build required exact CUDA–PyTorch version matching and ∼20 min of compilation. Bottom: cuRoboV2 installs via pip; kernels are compiled and cached on first use with no version constraints. cuRoboV2 replaces pybind11 with NVIDIA’s CUDA Python library [52], which compiles kernels at first use, automatically targets the host GPU, and caches results by sourc… view at source ↗
Figure 31
Figure 31. Figure 31: The RobotBuilder pipeline converts a URDF into a complete, optimized robot configuration through four automated stages. cuRoboV1 required users to manually author YAML configuration files specifying collision sphere place￾ments, self-collision ignore matrices, and joint-space parameters. This was error-prone: there was no au￾tomated way to determine which link pairs could safely be ignored, leading to con… view at source ↗
Figure 32
Figure 32. Figure 32: Camera-to-robot calibration via ICP. The user specifies a bounding box in the scene (left), which is used to run global ICP, aligning the robot mesh to the observed point cloud (right). Robot Segmentation To remove the robot from depth images, we use our forward kinematics to position the collision spheres at the current joint state, project depth pixels to 3D using the calibrated extrinsics and camera in… view at source ↗
Figure 33
Figure 33. Figure 33: Composable cost manager. Each BaseCost subclass is registered by name and evaluated on its own CUDA stream. Bold box indicates a user-defined custom cost. The split between CUDA C++ and Warp is driven by stability versus modifiability. Forward kinemat￾ics, L-BFGS optimization, B-spline evaluation, and self-collision are mature algorithms that change infre￾quently; these remain as hand-tuned CUDA C++ kerne… view at source ↗
read the original abstract

Effective robot autonomy requires motion generation that is safe, feasible, and reactive. Current methods are fragmented: fast planners output physically unexecutable trajectories, reactive controllers struggle with high-fidelity perception, and existing solvers fail on high-DoF systems. We present cuRoboV2, a unified framework with three key innovations: (1) B-spline trajectory optimization that enforces smoothness and torque limits; (2) a GPU-native TSDF/ESDF perception pipeline that generates dense signed distance fields covering the full workspace, unlike existing methods that only provide distances within sparsely allocated blocks, up to 10x faster and in 8x less memory than the state-of-the-art at manipulation scale, with up to 99% collision recall; and (3) scalable GPU-native whole-body computation, namely topology-aware kinematics, differentiable inverse dynamics, and map-reduce self-collision, that achieves up to 61x speedup while also extending to high-DoF humanoids (where previous GPU implementations fail). On benchmarks, cuRoboV2 achieves 99.7% success under 3kg payload (where baselines achieve only 72--77%), 99.6% collision-free IK on a 48-DoF humanoid (where prior methods fail entirely), and 89.5% retargeting constraint satisfaction (vs. 61% for PyRoki); these collision-free motions yield locomotion policies with 21% lower tracking error than PyRoki and 12x lower cross-seed variance than GMR. A ground-up codebase redesign for discoverability enabled LLM coding assistants to author up to 73% of new modules, including hand-optimized CUDA kernels, demonstrating that well-structured robotics code can unlock productive human-LLM collaboration. Together, these advances provide a unified, dynamics-aware motion generation stack that scales from single-arm manipulators to full humanoids. Code is available at https://github.com/NVlabs/curobo.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents cuRoboV2, a unified GPU-accelerated framework for dynamics-aware motion generation on high-DoF robots. Key contributions include B-spline trajectory optimization enforcing smoothness and torque limits, a GPU-native TSDF/ESDF perception pipeline producing dense signed distance fields (claimed up to 10x faster and 8x less memory than prior methods with up to 99% collision recall), and scalable whole-body solvers (topology-aware kinematics, differentiable inverse dynamics, map-reduce self-collision) achieving up to 61x speedup and extending to 48-DoF humanoids. Benchmark results report 99.7% success under 3 kg payload (vs. 72-77% baselines), 99.6% collision-free IK on 48-DoF humanoid (where priors fail), 89.5% retargeting satisfaction (vs. 61% for PyRoki), and downstream locomotion improvements; the work also highlights a codebase redesign enabling LLM assistance for up to 73% of new modules.

Significance. If the empirical gains hold under broader conditions, the integrated stack could meaningfully advance scalable, reactive motion planning for complex robots by tightly coupling perception, dynamics, and optimization in a GPU-native setting. The open code release and documented LLM-assisted development process add value for reproducibility and community adoption.

major comments (2)
  1. [Abstract] Abstract and benchmark results: the reported metrics (99.7% success under 3 kg payload, 99.6% collision-free IK, 89.5% retargeting) are given without error bars, trial counts, variance across seeds, or details on hyper-parameter selection and data exclusion, making it difficult to judge whether the gains over baselines are statistically robust or sensitive to tuning.
  2. [Abstract and Results] Perception and whole-body pipeline: the central claim of enabling safe, reactive autonomy in unstructured environments rests on the TSDF/ESDF pipeline and differentiable solvers remaining accurate and real-time under sensor noise, calibration drift, and dynamic obstacles, yet all quantitative results appear confined to controlled simulation benchmarks with no transfer experiments or noise-injection tests.
minor comments (2)
  1. [Abstract] The abstract states 'up to 99% collision recall' without specifying the exact evaluation protocol, scene types, or comparison methods used to obtain this figure.
  2. [Methods] B-spline optimization weights are identified as free parameters; if the manuscript claims any component is parameter-free, this should be clarified against the listed weights.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments point-by-point below. We have revised the manuscript to improve statistical reporting and to clarify the scope and limitations of the presented experiments.

read point-by-point responses

  1. Referee: [Abstract] Abstract and benchmark results: the reported metrics (99.7% success under 3 kg payload, 99.6% collision-free IK, 89.5% retargeting) are given without error bars, trial counts, variance across seeds, or details on hyper-parameter selection and data exclusion, making it difficult to judge whether the gains over baselines are statistically robust or sensitive to tuning.

    Authors: We agree that additional statistical details strengthen the claims. In the revised manuscript we have expanded the Experiments section to report: (i) number of trials (N=100 per benchmark), (ii) mean and standard deviation across five random seeds, (iii) error bars on all bar plots, and (iv) the hyper-parameter selection protocol together with the (very small) set of excluded trials due to solver timeouts. These additions are now also referenced from the abstract. revision: yes

  2. Referee: [Abstract and Results] Perception and whole-body pipeline: the central claim of enabling safe, reactive autonomy in unstructured environments rests on the TSDF/ESDF pipeline and differentiable solvers remaining accurate and real-time under sensor noise, calibration drift, and dynamic obstacles, yet all quantitative results appear confined to controlled simulation benchmarks with no transfer experiments or noise-injection tests.

    Authors: We acknowledge that the quantitative results are obtained in simulation. The manuscript’s primary contribution is the algorithmic unification and scaling behavior under controlled conditions that isolate the effect of each component. We have added a dedicated Limitations paragraph in the revised manuscript that explicitly states the absence of real-world transfer experiments and noise-injection studies, and we outline planned follow-up work. The simulation environments already incorporate sensor noise models; we have expanded the description of these models and added a short ablation on increasing noise levels. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical benchmarks against external baselines

full rationale

The paper introduces algorithmic components (B-spline trajectory optimization, GPU-native TSDF/ESDF pipeline, topology-aware kinematics, differentiable inverse dynamics, map-reduce self-collision) and reports success rates, IK success, and retargeting metrics via direct comparison to external baselines (PyRoki, GMR). No equations, predictions, or central claims reduce by construction to fitted parameters, self-defined quantities, or load-bearing self-citations. All quantitative results are presented as measured outcomes on benchmarks rather than derived identities. This matches the default expectation of a self-contained empirical contribution with no detectable circular steps.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The framework rests on standard assumptions about B-spline smoothness, signed distance field accuracy from depth fusion, and the validity of inverse dynamics models; no new physical entities are postulated.

free parameters (1)
  • B-spline optimization weights
    Weights balancing smoothness, torque limits, and collision costs are chosen to produce executable trajectories; exact values are not stated in the abstract.
axioms (2)
  • standard math B-splines can represent sufficiently smooth trajectories that satisfy torque limits when optimized
    Invoked in the first innovation for trajectory generation.
  • domain assumption Depth-fused TSDF/ESDF fields accurately represent workspace geometry at manipulation scale
    Central to the perception pipeline claim of 99% collision recall.

pith-pipeline@v0.9.0 · 5676 in / 1545 out tokens · 35994 ms · 2026-05-15T15:41:27.115091+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

138 extracted references · 138 canonical work pages · 2 internal anchors

  1. [1]

    Retargeting Matters: General motion retargeting for humanoid motion tracking.arXiv preprint arXiv:2510.02252, 2025

    Joao Pedro Araujo, Yanjie Ze, Pei Xu, Jiajun Wu, and C. Karen Liu. Retargeting matters: General motion retargeting for humanoid motion tracking. arXiv preprint arXiv:2510.02252, 2025. 6, 25

    work page arXiv 2025

  2. [2]

    Ratliff, Dieter Fox, Fabio Ramos, and Byron Boots

    Mohak Bhardwaj, Balakumar Sundaralingam, Arsalan Mousavian, Nathan D. Ratliff, Dieter Fox, Fabio Ramos, and Byron Boots. STORM: An Integrated Framework for Fast Joint-Space Model-Predictive Control for Reactive Manipulation. In Proceedings of the 5th Conference on Robot Learning, volume 164 of Proceedings of Machine Learning Research , pages 750–759. PMLR...

    work page 2022

  3. [3]

    Parallel banding algorithm to com- pute exact distance transform with the gpu

    Thanh-Tung Cao, Ke Tang, Anis Mohamed, and Tiow-Seng Tan. Parallel banding algorithm to com- pute exact distance transform with the gpu. In Proceedings of the 2010 ACM SIGGRAPH symposium on Interactive 3D Graphics and Games , pages 83–90, 2010. 12

    work page 2010

  4. [4]

    Analytical derivatives of rigid body dynamics algorithms

    Justin Carpentier and Nicolas Mansard. Analytical derivatives of rigid body dynamics algorithms. In Robotics: Science and systems (RSS 2018) , 2018. 18

    work page 2018

  5. [5]

    The pinocchio c++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives

    Justin Carpentier, Guilhem Saurel, Gabriele Buondonno, Joseph Mirabel, Florent Lamiraux, Olivier Stasse, and Nicolas Mansard. The pinocchio c++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives. In IEEE International Symposium on System Integrations (SII), 2019. 35

    work page 2019

  6. [6]

    Kingston, Carlos Quintero-Peña, Anshumali Shrivastava, and Ly- dia E

    Constantinos Chamzas, Zachary K. Kingston, Carlos Quintero-Peña, Anshumali Shrivastava, and Ly- dia E. Kavraki. Learning sampling distributions using local 3d workspace decompositions for motion planning in high dimensions. 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages 1283–1289, 2021. 4

    work page 2021

  7. [7]

    Motionbenchmaker: A tool to generate and benchmark motion planning datasets

    Constantinos Chamzas, Carlos Quintero-Pena, Zachary Kingston, Andreas Orthey, Daniel Rakita, Michael Gleicher, Marc Toussaint, and Lydia E Kavraki. Motionbenchmaker: A tool to generate and benchmark motion planning datasets. IEEE Robotics and Automation Letters , 7(2):882–889, 2022. 20

    work page 2022

  8. [8]

    Bodex: Scalable and efficient robotic dexterous grasp synthesis using bilevel optimization

    Jiayi Chen, Yubin Ke, and He Wang. Bodex: Scalable and efficient robotic dexterous grasp synthesis using bilevel optimization. In 2025 IEEE International Conference on Robotics and Automation (ICRA) , pages 01–08. IEEE, 2025. 5

    work page 2025

  9. [9]

    Rmp flow: A computational graph for automatic motion policy generation

    Ching-An Cheng, Mustafa Mukadam, Jan Issac, Stan Birchfield, Dieter Fox, Byron Boots, and Nathan Ratliff. Rmp flow: A computational graph for automatic motion policy generation. In International Workshop on the Algorithmic Foundations of Robotics , pages 441–457. Springer, 2018. 3, 4

    work page 2018

  10. [10]

    Physical-consistent behavior embodied in B-spline curves for robot path planning.IFAC-PapersOnLine, 51(22):306–311, 2018

    Davide Chiaravalli, Federico Califano, Luigi Biagiotti, Daniele De Gregorio, and Claudio Melchiorri. Physical-consistent behavior embodied in B-spline curves for robot path planning.IFAC-PapersOnLine, 51(22):306–311, 2018. 4

    work page 2018

  11. [12]

    Neural mp: A generalist neural motion planner

    Murtaza Dalal, Jiahui Yang, Russell Mendonca, Youssef Khaky, Ruslan Salakhutdinov, and Deepak Pathak. Neural mp: A generalist neural motion planner. arXiv preprint arXiv:2409.05864, 2024. 4, 6

    work page arXiv 2024

  12. [13]

    Object rearrangement us- ing learned implicit collision functions

    Michael Danielczuk, Arsalan Mousavian, Clemens Eppner, and Dieter Fox. Object rearrangement us- ing learned implicit collision functions. 2021 IEEE International Conference on Robotics and Automation (ICRA), pages 6010–6017, 2021. 6 40 cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots

    work page 2021

  13. [14]

    Object rearrangement using learned implicit collision functions

    Michael Danielczuk, Arsalan Mousavian, Clemens Eppner, and Dieter Fox. Object rearrangement using learned implicit collision functions. In 2021 IEEE International Conference on Robotics and Au- tomation (ICRA), pages 6010–6017. IEEE, 2021. 3

    work page 2021

  14. [15]

    Robot dynamics: equations and algorithms

    Roy Featherstone and David Orin. Robot dynamics: equations and algorithms. In Proceedings 2000 ICRA. Millennium conference. IEEE international conference on robotics and automation. Symposia pro- ceedings (Cat. No. 00CH37065) , volume 1, pages 826–834. IEEE, 2000. 16, 19

    work page 2000

  15. [16]

    Motion policy networks

    Adam Fishman, Adithyavairavan Murali, Clemens Eppner, Bryan Peele, Byron Boots, and Dieter Fox. Motion policy networks. In Proceedings of the 6th Conference on Robot Learning (CoRL) , 2022. 3, 4, 6, 20

    work page 2022

  16. [17]

    Avoid everything: Model-free collision avoidance with expert-guided fine-tuning

    Adam Fishman, Aaron Walsman, Mohak Bhardwaj, Wentao Yuan, Balakumar Sundaralingam, By- ron Boots, and Dieter Fox. Avoid everything: Model-free collision avoidance with expert-guided fine-tuning. In Proceedings of The 8th Conference on Robot Learning , volume 270 of Proceedings of Ma- chine Learning Research, pages 1925–1948. PMLR, 2025. URL https://procee...

    work page 1925

  17. [18]

    Humanplus: Humanoid shadowing and imitation from humans

    Zipeng Fu, Qingqing Zhao, Qi Wu, Gordon Wetzstein, and Chelsea Finn. Humanplus: Humanoid shadowing and imitation from humans. In Proceedings of The 8th Conference on Robot Learning, volume 270 of Proceedings of Machine Learning Research , pages 2828–2844. PMLR, 2025. 6

    work page 2025

  18. [19]

    Gammell, Siddhartha S

    Jonathan D. Gammell, Siddhartha S. Srinivasa, and Tim D. Barfoot. Batch informed trees (bit*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs. 2015 IEEE International Conference on Robotics and Automation (ICRA) , pages 3067–3074, 2015. 4

    work page 2015

  19. [20]

    FIESTA: Fast incremental euclidean distance fields for online motion planning of aerial robots

    Luxin Han, Fei Gao, Boyu Zhou, and Shaojie Shen. FIESTA: Fast incremental euclidean distance fields for online motion planning of aerial robots. In 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 4423–4430. IEEE, 2019. doi: 10.1109/IROS40897.2019.8968200. 6

    work page doi:10.1109/iros40897.2019.8968200 2019

  20. [21]

    Hart, Nils J

    Peter E. Hart, Nils J. Nilsson, and Bertram Raphael. A formal basis for the heuristic determination of minimum cost paths. IEEE Trans. Syst. Sci. Cybern. , 4:100–107, 1968. 4

    work page 1968

  21. [22]

    Learn- ing human-to-humanoid real-time whole-body teleoperation

    Tairan He, Zhengyi Luo, Wenli Xiao, Chong Zhang, Kris Kitani, Changliu Liu, and Guanya Shi. Learn- ing human-to-humanoid real-time whole-body teleoperation. In 2024 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) , pages 8944–8951. IEEE, 2024. 6

    work page 2024

  22. [23]

    Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoper- ation and learning

    Tairan He, Zhengyi Luo, Xialin He, Wenli Xiao, Chong Zhang, Weinan Zhang, Kris M Kitani, Changliu Liu, and Guanya Shi. Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoper- ation and learning. In Conference on Robot Learning , pages 1516–1540. PMLR, 2025. 6

    work page 2025

  23. [24]

    Diffusionseeder: Seeding motion optimization with diffusion for rapid motion planning

    Huang Huang, Balakumar Sundaralingam, Arsalan Mousavian, Adithyavairavan Murali, Ken Gold- berg, and Dieter Fox. Diffusionseeder: Seeding motion optimization with diffusion for rapid motion planning. In 8th Annual Conference on Robot Learning (CoRL), 2024. URL https://diffusion-seeder. github.io/. 4, 6

    work page 2024

  24. [25]

    Hydra: Marker-free rgb-d hand-eye calibration

    Martin Huber, Huanyu Tian, Christopher E Mower, Lucas-Raphael Müller, Sébastien Ourselin, Chris- tos Bergeles, and Tom Vercauteren. Hydra: Marker-free rgb-d hand-eye calibration. arXiv preprint arXiv:2504.20584, 2025. 48

    work page arXiv 2025

  25. [26]

    Spencer Williams, and Marc D

    Philip Hyatt, C. Spencer Williams, and Marc D. Killpack. Parameterized and GPU-parallelized real- time model predictive control for high degree-of-freedom robots, 2020. 4 41 cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots

    work page 2020

  26. [27]

    Deep learning can accelerate grasp- optimized motion planning

    Jeffrey Ichnowski, Yahav Avigal, Vishal Satish, and Ken Goldberg. Deep learning can accelerate grasp- optimized motion planning. Science Robotics, 5(48):eabd7710, 2020. doi: 10.1126/scirobotics.abd7710. 4

    work page doi:10.1126/scirobotics.abd7710 2020

  27. [28]

    Group marching tree: Sampling-based approx- imately optimal motion planning on GPUs

    Brian Ichter, Edward Schmerling, and Marco Pavone. Group marching tree: Sampling-based approx- imately optimal motion planning on GPUs. In IEEE International Conference on Robotic Computing (IRC), pages 219–226. IEEE, 2017. 4

    work page 2017

  28. [29]

    Learning sampling distributions for robot motion planning

    Brian Ichter, James Harrison, and Marco Pavone. Learning sampling distributions for robot motion planning. 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages 7087–7094,

    work page 2018

  29. [30]

    A convex optimization method to time-optimal trajectory planning with jerk constraint for industrial robotic manipulators

    Chen Ji, Zhongqiang Zhang, Guanggui Cheng, Minxiu Kong, and Ruifeng Li. A convex optimization method to time-optimal trajectory planning with jerk constraint for industrial robotic manipulators. IEEE Transactions on Automation science and engineering , 21(4):7629–7646, 2023. 4

    work page 2023

  30. [31]

    Exbody2: Advanced expres- sive humanoid whole-body control

    Mazeyu Ji, Xuanbin Guo, Xiaoyu He, Xue Bin Peng, and Xiaolong Wang. Exbody2: Advanced expres- sive humanoid whole-body control. arXiv preprint arXiv:2412.13196, 2024. 6

    work page arXiv 2024

  31. [32]

    Stomp: Stochastic trajectory optimization for motion planning

    Mrinal Kalakrishnan, Sachin Chitta, Evangelos Theodorou, Peter Pastor, and Stefan Schaal. Stomp: Stochastic trajectory optimization for motion planning. In 2011 IEEE international conference on robotics and automation , pages 4569–4574. IEEE, 2011. 4

    work page 2011

  32. [33]

    Sampling-based algorithms for optimal motion planning

    Sertaç Karaman and Emilio Frazzoli. Sampling-based algorithms for optimal motion planning. The International Journal of Robotics Research , 30:846 – 894, 2011. 4

    work page 2011

  33. [34]

    Neural collision clearance estimator for batched motion planning

    J Chase Kew, Brian Ichter, Maryam Bandari, Tsang-Wei Edward Lee, and Aleksandra Faust. Neural collision clearance estimator for batched motion planning. arXiv preprint arXiv:1910.05917, 2019. 6

    work page arXiv 1910

  34. [35]

    Real-time obstacle avoidance for manipulators and mobile robots

    Oussama Khatib. Real-time obstacle avoidance for manipulators and mobile robots. In Autonomous robot vehicles, pages 396–404. Springer, 1986. 4

    work page 1986

  35. [36]

    PyRoki: A modular toolkit for robot kinematic optimization

    Chung Min Kim, Brent Yi, Hongsuk Choi, Yi Ma, Ken Goldberg, and Angjoo Kanazawa. PyRoki: A modular toolkit for robot kinematic optimization. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2025. URL https://arxiv.org/abs/2505.03728. 5, 6, 23

    work page arXiv 2025

  36. [37]

    A probabilistic B-spline motion planning algorithm for un- manned helicopters flying in dense 3D environments

    Emre Koyuncu and Gokhan Inalhan. A probabilistic B-spline motion planning algorithm for un- manned helicopters flying in dense 3D environments. In 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 815–820. IEEE, 2008. 4

    work page 2008

  37. [38]

    Roboballet: Planning for multi-robot reaching with graph neural networks and reinforcement learn- ing

    Matthew Lai, Keegan Go, Zhibin Li, Torsten Kröger, Stefan Schaal, Kelsey Allen, and Jonathan Scholz. Roboballet: Planning for multi-robot reaching with graph neural networks and reinforcement learn- ing. Science Robotics, 10(106):eads1204, 2025. doi: 10.1126/scirobotics.ads1204. 4

    work page doi:10.1126/scirobotics.ads1204 2025

  38. [39]

    Steven M. LaValle. Rapidly-exploring random trees : a new tool for path planning. The annual research report, 1998. 4

    work page 1998

  39. [40]

    Gordon, and Sebastian Thrun

    Maxim Likhachev, Geoffrey J. Gordon, and Sebastian Thrun. Ara*: Anytime a* with provable bounds on sub-optimality. In NIPS, 2003. 4

    work page 2003

  40. [41]

    Warp: A high-performance python framework for gpu simulation and graphics.https: //github.com/nvidia/warp, March 2022

    Miles Macklin. Warp: A high-performance python framework for gpu simulation and graphics.https: //github.com/nvidia/warp, March 2022. NVIDIA GPU Technology Conference (GTC). 16, 38, 49

    work page 2022

  41. [42]

    A linear time algorithm for computing exact euclidean distance transforms of binary images in arbitrary dimensions

    Calvin R Maurer, Rensheng Qi, and Vijay Raghavan. A linear time algorithm for computing exact euclidean distance transforms of binary images in arbitrary dimensions. IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(2):265–270, 2003. 12 42 cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots

    work page 2003

  42. [43]

    Srmp: Search- based robot motion planning library

    Itamar Mishani, Yorai Shaoul, Ramkumar Natarajan, Jiaoyang Li, and Maxim Likhachev. Srmp: Search- based robot motion planning library. arXiv preprint arXiv:2509.25352, 2025. 3, 4

    work page arXiv 2025

  43. [44]

    Cabi- net: Scaling neural collision detection for object rearrangement with procedural scene generation

    Adithyavairavan Murali, Arsalan Mousavian, Clemens Eppner, Adam Fishman, and Dieter Fox. Cabi- net: Scaling neural collision detection for object rearrangement with procedural scene generation. In International Conference on Robotics and Automation (ICRA) , 2023. 3, 6

    work page 2023

  44. [45]

    Robot Motion Planning on a Chip

    Sean Murray, Will Floyd-Jones, Ying Qi, Daniel Sorin, and George Konidaris. Robot Motion Planning on a Chip. In Robotics Science and Systems , 2016. 4

    work page 2016

  45. [46]

    MorphIt: Flexible Spherical Approximation of Robot Morphology for Representation-driven Adaptation

    Nataliya Nechyporenko, Yutong Zhang, Sean Campbell, and Alessandro Roncone. Morphit: Flexible spherical approximation of robot morphology for representation-driven adaptation, 2025. URL https: //arxiv.org/abs/2507.14061. 47

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  46. [47]

    Neuman, Brian Plancher, Thomas Bourgeat, Thierry Tambe, Srinivas Devadas, and Vi- jay Janapa Reddi

    Sabrina M. Neuman, Brian Plancher, Thomas Bourgeat, Thierry Tambe, Srinivas Devadas, and Vi- jay Janapa Reddi. Robomorphic computing: A design methodology for domain-specific accelerators parameterized by robot morphology. In ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) , pages 674–686. ACM...

    work page 2021

  47. [48]

    Davison, Pushmeet Kohli, Jamie Shotton, Steve Hodges, and Andrew Fitzgibbon

    Richard Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J. Davison, Pushmeet Kohli, Jamie Shotton, Steve Hodges, and Andrew Fitzgibbon. Kinectfusion: Real-time dense surface mapping and tracking. 2011 10th IEEE International Symposium on Mixed and Augmented Reality (ISMAR), 2011. 6

    work page 2011

  48. [49]

    Newton: GPU-accelerated physics simulation for robotics, and simulation research, 2025

    Newton Contributors. Newton: GPU-accelerated physics simulation for robotics, and simulation research, 2025. URL https://github.com/newton-physics/newton. 16, 18, 23

    work page 2025

  49. [50]

    Real-time 3d reconstruc- tion at scale using voxel hashing

    Matthias Nießner, Michael Zollhöfer, Shahram Izadi, and Marc Stamminger. Real-time 3d reconstruc- tion at scale using voxel hashing. ACM Transactions on Graphics (ToG), 32(6):1–11, 2013. 9

    work page 2013

  50. [51]

    GitHub - nvidia-isaac/nvblox: A GPU-accelerated TSDF and ESDF library for robots equipped with RGB-D cameras

    NVIDIA. GitHub - nvidia-isaac/nvblox: A GPU-accelerated TSDF and ESDF library for robots equipped with RGB-D cameras. https://github.com/nvidia-isaac/nvblox, 2022. Accessed: 2022- 09-14. 3, 6, 30

    work page 2022

  51. [52]

    Cuda python, 2025

    NVIDIA. Cuda python, 2025. URL https://nvidia.github.io/cuda-python/latest/index.html. 46

    work page 2025

  52. [53]

    Voxblox: Incremen- tal 3d euclidean signed distance fields for on-board mav planning

    Helen Oleynikova, Zachary Taylor, Marius Fehr, Roland Siegwart, and Juan Nieto. Voxblox: Incremen- tal 3d euclidean signed distance fields for on-board mav planning. InIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2017. 6

    work page 2017

  53. [54]

    Collision-free and curvature-continuous path smooth- ing in cluttered environments

    Jia Pan, Liangjun Zhang, and Dinesh Manocha. Collision-free and curvature-continuous path smooth- ing in cluttered environments. In Robotics: Science and Systems (RSS) , 2011. URL https://www. roboticsproceedings.org/rss07/p32.pdf. 4

    work page 2011

  54. [55]

    Colored point cloud registration revisited

    Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun. Colored point cloud registration revisited. In ICCV,

    work page

  55. [56]

    Mimickit: A reinforcement learning framework for motion imitation and control

    Xue Bin Peng. Mimickit: A reinforcement learning framework for motion imitation and control. 2025. URL https://arxiv.org/abs/2510.13794. 26

    work page arXiv 2025

  56. [57]

    Deepmimic: Example- guided deep reinforcement learning of physics-based character skills

    Xue Bin Peng, Pieter Abbeel, Sergey Levine, and Michiel van de Panne. Deepmimic: Example- guided deep reinforcement learning of physics-based character skills. ACM Trans. Graph., 37(4):143:1– 143:14, July 2018. ISSN 0730-0301. doi: 10.1145/3197517.3201311. URL http://doi.acm.org/10.1145/ 3197517.3201311. 26 43 cuRoboV2: Dynamics-Aware Motion Generation wi...

    work page doi:10.1145/3197517.3201311 2018

  57. [58]

    GRiD: GPU-accelerated rigid body dynamics with analytical gradients

    Brian Plancher, Sabrina M Neuman, Radhika Ghosal, Scott Kuindersma, and Vijay Janapa Reddi. GRiD: GPU-accelerated rigid body dynamics with analytical gradients. In 2022 International Conference on Robotics and Automation (ICRA) , pages 6253–6260. IEEE, 2022. 18, 23

    work page 2022

  58. [59]

    C. Qi, L. Yi, Hao Su, and Leonidas J. Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. In NIPS, 2017. 6

    work page 2017

  59. [60]

    Ramsey, Zachary Kingston, Wil Thomason, and Lydia E

    Clayton W. Ramsey, Zachary Kingston, Wil Thomason, and Lydia E. Kavraki. Collision-affording point trees: SIMD-amenable nearest neighbors for fast collision checking. In Robotics: Science and Systems ,

    work page

  60. [61]

    URL http://arxiv.org/abs/2406.02807

    doi: 10.15607/RSS.2024.XX.038. URL http://arxiv.org/abs/2406.02807. 3, 9

    work page doi:10.15607/rss.2024.xx.038 2024

  61. [62]

    Chomp: Gradient optimiza- tion techniques for efficient motion planning

    Nathan Ratliff, Matt Zucker, J Andrew Bagnell, and Siddhartha Srinivasa. Chomp: Gradient optimiza- tion techniques for efficient motion planning. In 2009 IEEE International Conference on Robotics and Automation, pages 489–494. IEEE, 2009. 4, 7, 13

    work page 2009

  62. [63]

    Riemannian Motion Policies

    Nathan D. Ratliff, Jan Issac, Daniel Kappler, Stan Birchfield, and Dieter Fox. Riemannian motion policies. ArXiv, abs/1801.02854, 2018. 3, 4

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  63. [64]

    Jump flooding in gpu with applications to voronoi diagram and distance transform

    Guodong Rong and Tiow-Seng Tan. Jump flooding in gpu with applications to voronoi diagram and distance transform. In Proceedings of the 2006 ACM SIGGRAPH symposium on Interactive 3D Graphics and Games, pages 109–116, 2006. 12

    work page 2006

  64. [65]

    DART: Dense Articulated Real-time Tracking with consumer depth cameras

    Tanner Schmidt, Richard Newcombe, and Dieter Fox. DART: Dense Articulated Real-time Tracking with consumer depth cameras. Autonomous Robots, 39(3):239–258, 2015. ISSN 0929-5593. doi: 10.1007/ s10514-015-9462-z. URL http://dx.doi.org/10.1007/s10514-015-9462-z . 48

    work page doi:10.1007/s10514-015-9462-z 2015

  65. [66]

    Motion planning with sequential convex optimization and convex collision checking

    John Schulman, Yan Duan, Jonathan Ho, Alex Lee, Ibrahim Awwal, Henry Bradlow, Jia Pan, Sachin Patil, Ken Goldberg, and Pieter Abbeel. Motion planning with sequential convex optimization and convex collision checking. The International Journal of Robotics Research , 33(9):1251–1270, 2014. 7

    work page 2014

  66. [67]

    Lee, Ibrahim Awwal, Henry Bradlow, Jia Pan, Sachin Patil, Ken Goldberg, and P

    John Schulman, Yan Duan, Jonathan Ho, Alex X. Lee, Ibrahim Awwal, Henry Bradlow, Jia Pan, Sachin Patil, Ken Goldberg, and P. Abbeel. Motion planning with sequential convex optimization and convex collision checking. The International Journal of Robotics Research , 33:1251 – 1270, 2014. 4

    work page 2014

  67. [68]

    Curobo: Parallelized collision-free robot motion generation

    Balakumar Sundaralingam, Siva Kumar Sastry Hari, Adam Fishman, Caelan Garrett, Karl Van Wyk, Valts Blukis, Alexander Millane, Helen Oleynikova, Ankur Handa, Fabio Ramos, Nathan Ratliff, and Dieter Fox. Curobo: Parallelized collision-free robot motion generation. In 2023 IEEE International Conference on Robotics and Automation (ICRA) , pages 8112–8119, 202...

    work page doi:10.1109/icra48891.2023 2023

  68. [69]

    Drake: Model-based design and verification for robotics, 2019

    Russ Tedrake and the Drake Development Team. Drake: Model-based design and verification for robotics, 2019. URL https://drake.mit.edu. 4

    work page 2019

  69. [70]

    Wil Thomason*, Zachary Kingston*, and Lydia E. Kavraki. Motions in microseconds via vectorized sampling-based planning. In arxiv, 2023. 3, 4, 9, 20

    work page 2023

  70. [71]

    A survey of FPGA-based robotic computing

    Zishen Wan, Bo Yu, Tsung Yi Li, Juntao Tang, Yanjun Zhu, Yu Wang, Arijit Raychowdhury, and Shaoshan Liu. A survey of FPGA-based robotic computing. IEEE Circuits and Systems Magazine , 21 (2):48–74, 2021. 4

    work page 2021

  71. [72]

    Theodorou

    Grady Williams, Andrew Aldrich, and Evangelos A. Theodorou. Model predictive path integral control: From theory to parallel computation. Journal of Guidance, Control, and Dynamics, 40(2):344–357, 2017. doi: 10.2514/1.G001921. 3, 4 44 cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots

    work page doi:10.2514/1.g001921 2017

  72. [73]

    Wilson, Wil Thomason, Zachary Kingston, and Jonathan D

    Tyler S. Wilson, Wil Thomason, Zachary Kingston, and Jonathan D. Gammell. AORRTC: Almost- surely asymptotically optimal planning with RRT-Connect. IEEE Robotics and Automation Letters ,

    work page

  73. [74]

    Under Review

    URL https://arxiv.org/abs/2505.10542. Under Review. 20

    work page arXiv

  74. [75]

    Peele, Qian Wan, Iretiayo Akinola, Balakumar Sundaralingam, Dieter Fox, Byron Boots, and Nathan D

    Karl Van Wyk, Mandy Xie, Anqi Li, Muhammad Asif Rana, Buck Babich, Bryan N. Peele, Qian Wan, Iretiayo Akinola, Balakumar Sundaralingam, Dieter Fox, Byron Boots, and Nathan D. Ratliff. Geomet- ric fabrics: Generalizing classical mechanics to capture the physics of behavior. IEEE Robotics and Automation Letters, 7:3202–3209, 2022. 4

    work page 2022

  75. [76]

    Deep reactive policy: Learning reactive manipulator motion planning for dynamic environments

    Jiahui Yang, Jason Jingzhou Liu, Yulong Li, Youssef Khaky, Kenneth Shaw, and Deepak Pathak. Deep reactive policy: Learning reactive manipulator motion planning for dynamic environments. In Pro- ceedings of the 9th Conference on Robot Learning (CoRL) , 2025. 4, 6

    work page 2025

  76. [77]

    Viser: Imperative, web-based 3d visualization in python, 2025

    Brent Yi, Chung Min Kim, Justin Kerr, Gina Wu, Rebecca Feng, Anthony Zhang, Jonas Kulhanek, Hongsuk Choi, Yi Ma, Matthew Tancik, and Angjoo Kanazawa. Viser: Imperative, web-based 3d visualization in python, 2025. URL https://arxiv.org/abs/2507.22885. 48

    work page arXiv 2025

  77. [78]

    Learning-based initialization of trajectory optimization for path-following problems of redundant manipulators

    Minsung Yoon, Mincheul Kang, Daehyung Park, and Sung-Eui Yoon. Learning-based initialization of trajectory optimization for path-following problems of redundant manipulators. In 2023 IEEE Interna- tional Conference on Robotics and Automation (ICRA) , pages 7670–7676. IEEE, 2023. 4

    work page 2023

  78. [79]

    Mink: Python inverse kinematics based on MuJoCo, May 2025

    Kevin Zakka. Mink: Python inverse kinematics based on MuJoCo, May 2025. URL https://github. com/kevinzakka/mink. 6 45 cuRoboV2: Dynamics-Aware Motion Generation with Depth-Fused Distance Fields for High-DoF Robots A. Framework Design Beyond the algorithmic contributions presented in previous sections, cuRoboV2 introduces substantial ar- chitectural change...

    work page 2025

  79. [80]

    Neighbor pairs: adjacent links in the kinematic chain are unconditionally ignored

    work page

  80. [81]

    Default configuration: pairs that collide at the retract pose are ignored, as they represent perma- nent geometric overlaps that planning cannot resolve

    work page

Showing first 80 references.