pith. sign in

arxiv: 1906.11176 · v1 · pith:536V43Z7new · submitted 2019-06-26 · 💻 cs.RO · cs.CV· cs.LG

PyRep: Bringing V-REP to Deep Robot Learning

Stephen James , Marc Freese , Andrew J. Davison This is my paper

Pith reviewed 2026-05-25 15:24 UTC · model grok-4.3

classification 💻 cs.RO cs.CVcs.LG
keywords PyRepV-REProbot learningreinforcement learningsimulationAPIrendering engine
0
0 comments X

The pith

PyRep toolkit adds a simple API and up to 10000x speed boost to V-REP for robot learning.

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

The paper presents PyRep as a modified version of the V-REP simulator built specifically for robot learning research. It claims three concrete improvements over the prior setup: a simple and flexible Python API for robot control and scene manipulation, a new rendering engine, and performance increases reaching 10,000 times the speed of the previous Python Remote API. These changes are meant to support rapid prototyping of algorithms in reinforcement learning, imitation learning, state estimation, mapping, and computer vision. A sympathetic reader would see value in simulation tools that reduce the computational cost of running many training episodes or vision-based experiments.

Core claim

Through a series of modifications and additions to V-REP, PyRep provides a simple and flexible API for robot control and scene manipulation, a new rendering engine, and speed boosts upwards of 10,000x in comparison to the previous Python Remote API. With these improvements, PyRep becomes the ideal toolkit to facilitate rapid prototyping of learning algorithms in the areas of reinforcement learning, imitation learning, state estimation, mapping, and computer vision.

What carries the argument

The PyRep toolkit, consisting of modifications and additions to V-REP that deliver the API, rendering engine, and performance gains.

If this is right

  • Training loops for reinforcement learning policies can complete many more episodes in the same wall-clock time.
  • Scene manipulation and robot control code becomes shorter and easier to integrate with learning frameworks.
  • Computer vision and mapping experiments gain from the new rendering engine without changing the underlying simulator.
  • Imitation learning and state estimation prototypes can be iterated on more quickly due to reduced per-step computation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The speed gains could support larger batch sizes or more complex scenes during learning without changing hardware.
  • Existing V-REP scene files might transfer directly, lowering the effort needed to adopt the toolkit in ongoing projects.

Load-bearing premise

The modifications to V-REP will deliver the stated speed and usability gains consistently across hardware and tasks without introducing new limitations that reduce usefulness for the listed learning applications.

What would settle it

Measure simulation steps per second when running a standard reinforcement learning training loop on identical hardware and tasks in both PyRep and the original V-REP Python Remote API to check whether the 10,000x speedup holds.

Figures

Figures reproduced from arXiv: 1906.11176 by Andrew J. Davison, Marc Freese, Stephen James.

Figure 1
Figure 1. Figure 1: Example images of environments using the new renderer. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: PyRep API Example. Many more examples can be seen on the GitHub page. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
read the original abstract

PyRep is a toolkit for robot learning research, built on top of the virtual robotics experimentation platform (V-REP). Through a series of modifications and additions, we have created a tailored version of V-REP built with robot learning in mind. The new PyRep toolkit offers three improvements: (1) a simple and flexible API for robot control and scene manipulation, (2) a new rendering engine, and (3) speed boosts upwards of 10,000x in comparison to the previous Python Remote API. With these improvements, we believe PyRep is the ideal toolkit to facilitate rapid prototyping of learning algorithms in the areas of reinforcement learning, imitation learning, state estimation, mapping, and computer vision.

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

1 major / 0 minor

Summary. The paper introduces PyRep, a modified version of the V-REP simulator tailored for robot learning research. It claims three main improvements over the base platform and prior Python Remote API: (1) a simple and flexible API for robot control and scene manipulation, (2) a new rendering engine, and (3) speed increases of up to 10,000x. These changes are positioned to support rapid prototyping in reinforcement learning, imitation learning, state estimation, mapping, and computer vision.

Significance. If the reported API usability gains and speed improvements hold under independent verification, PyRep would provide a practical, open toolkit that lowers the barrier to simulation-based robot learning experiments. The explicit focus on learning workloads and the release of modified simulator code represent a concrete engineering contribution that could accelerate iteration in the listed application areas.

major comments (1)
  1. [Abstract] Abstract: The central claim of 'speed boosts upwards of 10,000x in comparison to the previous Python Remote API' is presented without any benchmark details, task descriptions, hardware specifications, timing methodology, or error analysis. Because this quantitative improvement is one of the three headline contributions and is load-bearing for the paper's utility argument, the absence of supporting measurements prevents evaluation of whether the figure is reproducible or task-dependent.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their feedback. We address the single major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of 'speed boosts upwards of 10,000x in comparison to the previous Python Remote API' is presented without any benchmark details, task descriptions, hardware specifications, timing methodology, or error analysis. Because this quantitative improvement is one of the three headline contributions and is load-bearing for the paper's utility argument, the absence of supporting measurements prevents evaluation of whether the figure is reproducible or task-dependent.

    Authors: We agree the abstract would benefit from additional context on the speed claim. The full manuscript details the benchmarks (including task descriptions such as repeated forward kinematics and scene stepping, hardware specifications, timing methodology using repeated trials with Python timing, and observed variability) in the Experiments section. We will revise the abstract to concisely reference the benchmark conditions under which the up-to-10,000x figure was measured, while preserving brevity. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a descriptive software release note for the PyRep toolkit. It contains no derivations, equations, fitted parameters, predictions, or self-referential claims that reduce to inputs by construction. All central claims concern API design, a rendering engine update, and empirical speed measurements presented as direct observations from the implemented changes, with no load-bearing self-citations or uniqueness theorems invoked. The derivation chain is empty by nature of the paper's genre.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a software toolkit description rather than a theoretical paper. No free parameters, axioms, or invented entities are introduced or required by the central claims in the abstract.

pith-pipeline@v0.9.0 · 5648 in / 1087 out tokens · 28212 ms · 2026-05-25T15:24:39.287614+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

11 extracted references · 11 canonical work pages

  1. [1]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”Nature, vol. 521, no. 7553, p. 436, 2015

    work page 2015

  2. [2]

    Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross- domain image matching,

    A. Zeng, S. Song, K.-T. Yu, E. Donlon, F. R. Hogan, M. Bauza, D. Ma, O. Taylor, M. Liu, E. Romo, et al., “Robotic pick-and-place of novel objects in clutter with multi-affordance grasping and cross- domain image matching,” International Conference on Robotics and Automation , 2018

    work page 2018

  3. [3]

    Cartman: The low-cost cartesian manipulator that won the amazon robotics challenge,

    D. Morrison, A. W. Tow, M. McTaggart, R. Smith, N. Kelly-Boxall, S. Wade-McCue, J. Erskine, R. Grinover, A. Gurman, T. Hunn, et al., “Cartman: The low-cost cartesian manipulator that won the amazon robotics challenge,” International Conference on Robotics and Automation , 2018

    work page 2018

  4. [4]

    Using simulation and domain adaptation to improve efficiency of deep robotic grasping,

    K. Bousmalis, A. Irpan, P. Wohlhart, Y . Bai, M. Kelcey, M. Kalakrishnan, L. Downs, J. Ibarz, P. Pastor, K. Konolige, et al. , “Using simulation and domain adaptation to improve efficiency of deep robotic grasping,” International Conference on Robotics and Automation , 2018

    work page 2018

  5. [5]

    Sim-to-real via sim-to-sim: Data-efficient robotic grasping via randomized-to- canonical adaptation networks,

    S. James, P. Wohlhart, M. Kalakrishnan, D. Kalashnikov, A. Irpan, J. Ibarz, S. Levine, R. Hadsell, and K. Bousmalis, “Sim-to-real via sim-to-sim: Data-efficient robotic grasping via randomized-to- canonical adaptation networks,” Conference on Computer Vision and Pattern Recognition, 2019

    work page 2019

  6. [6]

    Domain randomization for transferring deep neural networks from simulation to the real world,

    J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel, “Domain randomization for transferring deep neural networks from simulation to the real world,” International Conference on Intelligent Robots and Systems , 2017

    work page 2017

  7. [7]

    Transferring end-to-end visuomotor control from simulation to real world for a multi-stage task,

    S. James, A. J. Davison, and E. Johns, “Transferring end-to-end visuomotor control from simulation to real world for a multi-stage task,” Conference on Robot Learning, 2017

    work page 2017

  8. [8]

    Sim-to-real reinforcement learning for deformable object manipulation,

    J. Matas, S. James, and A. J. Davison, “Sim-to-real reinforcement learning for deformable object manipulation,” Conference on Robot Learning, 2018

    work page 2018

  9. [9]

    Bullet physics simulation,

    E. Coumans, “Bullet physics simulation,” in ACM SIGGRAPH 2015 Courses , SIGGRAPH ’15, (New York, NY , USA), ACM, 2015

    work page 2015

  10. [10]

    Mujoco: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,”International Conference on Intelligent Robots and Systems , 2012

    work page 2012

  11. [11]

    V-rep: A versatile and scalable robot simulation framework,

    E. Rohmer, S. P. Singh, and M. Freese, “V-rep: A versatile and scalable robot simulation framework,” International Conference on Intelligent Robots and Systems , 2013. 4

    work page 2013