Pith. sign in

REVIEW 3 major objections 5 minor 43 references

Chargax: A JAX Accelerated EV Charging Simulator

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Chargax runs EV-charging reinforcement learning on GPUs via JAX, cutting training from days to minutes.

desk verdict Useful and honest JAX-based EV charging environment with a plausible speedup that needs a reproducible benchmark before the exact factors are taken at face value. read the letter →

arxiv 2507.01522 v1 pith:5E3XCM4X submitted 2025-07-02 cs.LG cs.AIcs.SYeess.SY

Koen Ponse , Jan Felix Kleuker , Aske Plaat , Thomas Moerland This is my paper
classification cs.LGcs.AIcs.SYeess.SY
keywords electricvehiclechargingreinforcementlearningJAXGPUaccelerationstationsimulationPPOsustainableenergygymenvironment
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces Chargax, a reinforcement-learning environment for electric-vehicle charging written entirely in JAX so the whole simulator runs on a GPU. It claims this is the first such EV-charging environment in JAX and that it is over 100x faster than existing Gym-style EV charging simulators, with speedups reaching 1000x in vectorized settings. The practical point is that training a PPO agent, which previously took working days, can finish in under five minutes, making large-scale experiments feasible. Chargax is validated on real-data scenarios against naive charging baselines, showing profit gains and tunable tradeoffs with user satisfaction. If true, the paper bridges the gap between toy JAX RL environments and realistic energy-system benchmarks.

What carries the argument

The load-bearing mechanism is a tree representation of the charging station's electrical infrastructure, with the grid connection at the root, splitters, transformers, and cables at internal nodes, and charging ports (EVSEs) at the leaves. Each node carries a capacity and an efficiency coefficient, and the agent's action sets currents at each port, which are rescaled when subtree constraints are violated. Around this tree, the environment state is split into endogenous variables, which the agent's actions affect, and exogenous variables, which evolve from time-series data such as arrivals, prices, and user profiles. The entire transition function is written in JAX, so it runs on GPU, is vectorized across parallel environments, and avoids CPU-GPU data movement during PPO training.

What would settle it

Benchmark Chargax against EV2Gym, Chargym, and Sustaingym with equivalent vectorized implementations and identical scenario definitions, then check whether the wall-clock time for 100k PPO steps still shows a 100x gap; if an optimized multi-environment CPU baseline closes the gap to below 10x, the central speedup claim would be falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that a full EV charging station simulator, including car arrivals, battery state, charging curves, power constraints, and reward calculation, can be expressed as JAX transition functions and vectorized on a GPU without sacrificing realism. The authors report that a standard PPO implementation completes 100k environment steps in 0.65 seconds with 16 parallel environments, versus 87 to 1836 seconds for EV2Gym, Chargym, and Sustaingym under the same protocol. They interpret this as demonstrating that JAX-based environments are not limited to toy problems and that realistic energy applications can benefit from the same GPU-accelerated training loop. The paper further shows that a PPO agent trained in Chargax outperforms max-charging and random baselines on profit, and that adding weighted penalties can improve user satisfaction with only a modest profit loss.

Load-bearing premise

The 100x-1000x speedup claim rests on comparing Chargax against three existing environments run with Stable-Baselines3 PPO on the paper's chosen hardware, even though the authors note these environments simulate different behaviors; if the baselines are slower than they need to be or the scenarios differ too much, the speedup ratio is inflated.

Editorial extensions

If this is right

  • RL training for EV charging control can be iterated hundreds of times per day, allowing hyperparameter sweeps, multi-seed studies, and algorithm comparison on realistic data.
  • Chargax's modular tree architecture lets one codebase express different station layouts, including AC and DC charger mixes, batteries, and multiple splitters, so results can transfer across configurations.
  • The built-in exogenous data pipeline makes distribution-shift testing straightforward; the paper's 2022 Dutch price experiment suggests training on a crisis year can be worse than training on neighboring years.
  • Because the environment is JAX-native, Chargax can serve as a fast testbed for developing new RL algorithms for resource allocation under real-world constraints.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension is that other heavy CPU-based energy simulators, such as building control, microgrids, and fleet routing, could be re-architected in JAX with similar orders-of-magnitude gains wherever the transition can be expressed as array operations.
  • A direct head-to-head implementation of the baseline simulators in JAX would isolate how much of the speedup comes from JAX itself rather than from Chargax's specific design choices.
  • Chargax's design as a general resource-allocation testbed suggests it could also benchmark offline RL or multi-agent variants, where current toy JAX environments are not representative of real-world constraints.
  • The 2022 distribution-shift finding hints that practitioners should check whether extraordinary market episodes make training data less useful than calmer years, a caution that generalizes beyond EV charging.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces Chargax, a JAX-based reinforcement learning environment for EV charging stations. It formalizes the environment as an MDP with an explicit split between endogenous and exogenous state, represents station infrastructure as a tree of splitters/transformers with capacity constraints, and provides several reward functions (profit, user satisfaction, sustainability, etc.) and real-data scenarios. The empirical sections show PPO training in multiple scenarios, a distribution-shift study on price data, and a benchmark (Table 2) claiming speedups of more than 100x-1000x over three existing EV charging Gym environments. The authors also claim that Chargax is the first EV charging RL environment implemented in JAX.

Significance. If the claimed speedups hold, this is a practically valuable contribution: it would reduce RL training for EV charging from hours or days to minutes, enable GPU-based hyperparameter and algorithm experiments, and provide a modular, JAX-native platform for a realistic resource-allocation problem. The paper also has clear strengths: the environment design is described in detail (tree-based architecture, explicit exogenous/endogenous state, flexible reward components), several scenarios use real data, the PPO learning curves are reported over multiple seeds in Figure 4a, and the open-source release is a step toward reproducibility. However, the headline speedup claim is not yet fully supported by the benchmark as reported: Table 2 lacks error bars, baseline configurations, and a reproducibility script, and it compares environments with different semantics. The contribution is defensible, but the central quantitative claim needs to be made auditable before acceptance.

major comments (3)
  1. [Section 5, Table 2 and Figure 1] The benchmark protocol is not sufficiently specified to support the headline speedup claim. The paper does not report the baseline environments' configurations (number of EVSEs, timestep duration, scenario, episode length, arrival/user/car distributions), the exact software versions, the number of repetitions, or error bars. Because the comparison is per '100k environment steps,' and the environments differ in their internal time grids and station sizes, the reported per-step speedups may not translate to speedups per simulated charging day. The authors should provide a benchmark script and a detailed configuration table for all four environments, or report throughput on matched workloads (e.g., same number of charging sessions simulated).
  2. [Section 5, Table 2 and the accompanying discussion] The comparison conflates environment speed with RL implementation speed. The PPO rows compare Chargax with PureJaxRL against EV2Gym, Chargym, and Sustaingym with Stable-Baselines3, so the measured factors include differences in PPO implementations, not only environment transition speed. The random-action rows are a cleaner environment-only comparison, but they still involve non-identical environments; the authors themselves state that 'these environments are not identical and might simulate different behaviours' (Section 5, Table 2 discussion). The claim 'the significant differences in scale clearly demonstrate the advantages' is not a substitute for a controlled comparison. Please report raw environment step throughput separately from end-to-end PPO training time, or use the same RL algorithm stack for all environments.
  3. [Abstract and Section 5, Table 2] The abstract and contribution list state a speedup of 'over 100x-1000x,' but the numbers in Table 2 do not consistently support this range. The PPO(1) row reports speedups of 13x-176x, the random row 27x-1144x, and the PPO(16) row 134x-2820x. If the headline claim is meant to apply to typical training (PPO with vectorized environments), that should be stated precisely, and the corresponding range should be quoted. If the claim is meant to cover all rows, the lower bounds contradict the 'over 100x' wording. The authors should rephrase the claim to match the evidence or provide additional measurements under the claimed conditions.
minor comments (5)
  1. [Appendix D, Figures 6-11] The caption of Figure 6 says '16 chargers (10 DC, 5 AC)', which sums to 15, while the main text and Table 3 describe a 16-charger station with 10 DC and 6 AC. This inconsistency should be corrected in all affected captions.
  2. [Table 2 and Figure 4 captions] There are small typos: 'assesing' in the Table 2 caption and 'devidation' in the Figure 4 caption. These should be fixed.
  3. [Section 4, Equation (4)] The formal definition of exogenous/endogenous state via a 'maximal index m' is not fully precise: reordering state variables could make the factorization non-unique, and the maximal index is not characterized further. Please clarify how the ordering is chosen or give a coordinate-free definition of exogeneity (e.g., transition independence from actions and endogenous variables).
  4. [Table 1 and Section 5] The price profiles (NL, FR, DE), car distributions, and arrival frequency datasets are described as based on real data, but no data sources or licenses are cited. Providing provenance would improve reproducibility and trust in the realism claims.
  5. [Figure 5 and Appendix B] The evaluation protocol for the distribution-shift experiment is not fully specified: the text says agents are evaluated for 36500 episodes, but it is not stated whether the evaluation uses a deterministic policy, which data are used for arrivals and user profiles, and how the 10 seeds are aggregated. Please add these details.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central speedup and RL results are direct empirical measurements, and its one formal decomposition is a definitional factorization rather than a derived prediction.

full rationale

Chargax is a software artifact, and the paper's central claims are empirical: the 100x-1000x speedup in Table 2 and Figure 1 are measured wall-clock times for 100k environment steps, the PPO learning curves in Figures 4, 6-11 are trained and evaluated on data, and the distribution-shift experiment in Figure 5 evaluates agents on years not used in training. None of these results are fitted parameters renamed as predictions; they are direct benchmark outputs. The only formal derivation-like passage is the exogenous/endogenous state decomposition in Appendix A.1, Eq. 4, which states that 'it is always possible to find a maximal index m' such that the transition function factorizes, and then defines exogenous and endogenous state spaces from that factorization. This is a definitional taxonomy, not a derivation of a substantive result, so it cannot be circular. The self-citations to Ponse et al., 2024 are used only as motivational context for the value of simulation environments and for the observation that an exogenous/endogenous separation is often implicit; neither citation is load-bearing for the speedup claim or for the environment's mechanics. The acknowledged point in Section 5 that 'these environments are not identical and might simulate different behaviours' raises benchmark-fairness and reproducibility questions about the exact speedup magnitude, but that is a correctness/auditability concern, not a circularity concern. No equation-level reduction of a claimed result to its own inputs was found.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The simulator introduces no new physical entities or fitted constants. It relies on standard domain simplifications for EV charging behavior, all stated explicitly in the appendices.

assumptions (4)
  • domain assumption A car's maximum charging rate follows a piece-wise linear curve with a single transition point tau, shared across AC and DC chargers.
    Appendix A.1; adopted from Lee et al. 2020b, with discharging rate assumed to be vertical flip due to lack of data.
  • domain assumption Charging power is constant over each 5-minute timestep.
    Appendix A.2, 'we assume a constant charging rate over the full interval Delta t'.
  • domain assumption The station is an isolated power network; shared transformers and external grid constraints are ignored.
    Section 6 acknowledges this simplifying assumption.
  • domain assumption Arrivals follow a Poisson distribution with per-timestep means from bundled datasets.
    Appendix B.1, 'the actual number of cars arriving is drawn using a Poisson distribution.'

how reviews work

0 comments
Cite this review

Pith. "Pith review of Chargax: A JAX Accelerated EV Charging Simulator." pith.science (2026). https://pith.science/paper/5E3XCM4X

@misc{pith2026250701522,
  author       = {Pith},
  title        = {Pith review of: Chargax: A JAX Accelerated EV Charging Simulator},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5E3XCM4X}},
  note         = {Machine review of arXiv:2507.01522}
}
read the original abstract

Deep Reinforcement Learning can play a key role in addressing sustainable energy challenges. For instance, many grid systems are heavily congested, highlighting the urgent need to enhance operational efficiency. However, reinforcement learning approaches have traditionally been slow due to the high sample complexity and expensive simulation requirements. While recent works have effectively used GPUs to accelerate data generation by converting environments to JAX, these works have largely focussed on classical toy problems. This paper introduces Chargax, a JAX-based environment for realistic simulation of electric vehicle charging stations designed for accelerated training of RL agents. We validate our environment in a variety of scenarios based on real data, comparing reinforcement learning agents against baselines. Chargax delivers substantial computational performance improvements of over 100x-1000x over existing environments. Additionally, Chargax' modular architecture enables the representation of diverse real-world charging station configurations.

Figures

Figures reproduced from arXiv: 2507.01522 by the authors.

Figure 1
Figure 1. Comparison between Chargax and prior EV Gym Environments in seconds to com￾plete 100k training steps using PPO. See [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of the Chargax environment. The [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Trees representing different architectures: a) simplest scenario, one type of charger; b) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: In a) average episode rewards during training a PPO agent in the shopping scenario with [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: A PPO agent trained and tested on three separate years of Dutch electricity prices. For [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Results on our 4 bundled scenarios using [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Results on our 4 bundled scenarios using [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Results on our 4 bundled scenarios using [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Results on our 4 bundled scenarios using EU cars and 16 AC (11.5kW) chargers [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: Results on our 4 bundled scenarios using EU cars and 8 AC (11.5kW) and 8 DC (150kW) [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: Results on our 4 bundled scenarios using EU cars and 16 DC (150kW) chargers [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 29 canonical work pages

  1. [1]

    A Review of the Latest Trends in Technical and Economic Aspects of EV Charging Management

    Pegah Alaee, Julius Bems, and Amjad Anvari-Moghaddam . A Review of the Latest Trends in Technical and Economic Aspects of EV Charging Management . Energies, 16 0 (9): 0 3669, January 2023. ISSN 1996-1073. doi:10.3390/en16093669

    work page doi:10.3390/en16093669 2023

  2. [2]

    An Optimal Management for Charging and Discharging of Electric Vehicles in an Intelligent Parking Lot Considering Vehicle Owner 's Random Behaviors

    Mahyar Alinejad, Omid Rezaei, Ahad Kazemi, and Saeed Bagheri. An Optimal Management for Charging and Discharging of Electric Vehicles in an Intelligent Parking Lot Considering Vehicle Owner 's Random Behaviors . Journal of Energy Storage, 35: 0 102245, March 2021. ISSN 2352152X. doi:10.1016/j.est.2021.102245

    work page arXiv 2021

  3. [3]

    Ev-ecosim: A grid-aware co-simulation platform for the design and optimization of electric vehicle charging infrastructure

    Emmanuel Balogun, Elizabeth Buechler, Siddharth Bhela, Simona Onori, and Ram Rajagopal. Ev-ecosim: A grid-aware co-simulation platform for the design and optimization of electric vehicle charging infrastructure. IEEE Transactions on Smart Grid, 2023

    work page 2023

  4. [4]

    Midgley, Elshadai Tegegn, Tristan Kalloniatis, Omayma Mahjoub, Matthew Macfarlane, Andries P

    Clément Bonnet, Daniel Luo, Donal Byrne, Shikha Surana, Sasha Abramowitz, Paul Duckworth, Vincent Coyette, Laurence I. Midgley, Elshadai Tegegn, Tristan Kalloniatis, Omayma Mahjoub, Matthew Macfarlane, Andries P. Smit, Nathan Grinsztajn, Raphael Boige, Cemlyn N. Waters, Mohamed A. Mimouni, Ulrich A. Mbou Sob, Ruan de Kock, Siddarth Singh, Daniel Furelos-B...

    arXiv 2024

  5. [5]

    JAX : composable transformations of P ython+ N um P y programs, 2018

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake Vander P las, Skye Wanderman- M ilne, and Qiao Zhang. JAX : composable transformations of P ython+ N um P y programs, 2018. URL http://github.com/jax-ml/jax

    2018

  6. [6]

    Cañigueral

    M. Cañigueral. evsim: Electric vehicle charging sessions simulation, 2023. R package version 1.2.0. [Online]. Available: https://github.com/mcanigueral/evsim/

    work page 2023

  7. [7]

    Coordinate scheduling of electric vehicles in charging stations supported by microgrids

    Shuo Chang, Yugang Niu, and Tinggang Jia. Coordinate scheduling of electric vehicles in charging stations supported by microgrids. Electric Power Systems Research, 199: 0 107418, October 2021. ISSN 03787796. doi:10.1016/j.epsr.2021.107418

    work page arXiv 2021

  8. [8]

    A dynamic charging strategy with hybrid fast charging station for electric vehicles

    Onur Elma. A dynamic charging strategy with hybrid fast charging station for electric vehicles. Energy, 202: 0 117680, July 2020. ISSN 03605442. doi:10.1016/j.energy.2020.117680

    work page arXiv 2020

Show all 43 references
  1. [9]

    Charge speed, 2025

    Fastned. Charge speed, 2025. URL https://www.fastnedcharging.com/en/brands-overview. Accessed: 2025-02-14

    2025

  2. [10]

    Actor-critic continuous state reinforcement learning for wind-turbine control robust optimization

    Borja Fernandez-Gauna , Manuel Gra \ n a, Juan-Luis Osa-Amilibia , and Xabier Larrucea. Actor-critic continuous state reinforcement learning for wind-turbine control robust optimization. Information Sciences, 591: 0 365--380, April 2022. ISSN 00200255. doi:10.1016/j.ins.2022.01.047

    2022 doi

  3. [11]

    Daniel Freeman, Erik Frey, Anton Raichuk, Sertan Girgin, Igor Mordatch, and Olivier Bachem

    C. Daniel Freeman, Erik Frey, Anton Raichuk, Sertan Girgin, Igor Mordatch, and Olivier Bachem. Brax - a differentiable physics engine for large scale rigid body simulation, 2021. URL http://github.com/google/brax

    2021

  4. [12]

    Flax: A neural network library and ecosystem for JAX , 2024

    Jonathan Heek, Anselm Levskaya, Avital Oliver, Marvin Ritter, Bertrand Rondepierre, Andreas Steiner, and Marc van Zee. Flax: A neural network library and ecosystem for JAX , 2024. URL http://github.com/google/flax

    2024

  5. [13]

    Podracer architectures for scalable reinforcement learning

    Matteo Hessel, Manuel Kroiss, Aidan Clark, Iurii Kemaev, John Quan, Thomas Keck, Fabio Viola, and Hado van Hasselt. Podracer architectures for scalable reinforcement learning. arXiv preprint arXiv:2104.06272, 2021

    2021 arXiv

  6. [14]

    Shahid Hussain, Yun-Su Kim, Subhasis Thakur, and John G. Breslin. Optimization of Waiting Time for Electric Vehicles Using a Fuzzy Inference System . IEEE Transactions on Intelligent Transportation Systems, 23 0 (9): 0 15396--15407, September 2022. ISSN 1558-0016. doi:10.1109/...

    2022

  7. [15]

    Campbell, Konrad Czechowski, Dumitru Erhan, Chelsea Finn, Piotr Kozakowski, Sergey Levine, Afroz Mohiuddin, Ryan Sepassi, George Tucker, and Henryk Michalewski

    Lukasz Kaiser, Mohammad Babaeizadeh, Piotr Milos, Blazej Osinski, Roy H. Campbell, Konrad Czechowski, Dumitru Erhan, Chelsea Finn, Piotr Kozakowski, Sergey Levine, Afroz Mohiuddin, Ryan Sepassi, George Tucker, and Henryk Michalewski. Model- Based Reinforcement Learning for Ata...

    2024 arXiv

  8. [16]

    Chargym: An ev charging station model for controller benchmarking

    Georgios Karatzinis, Christos Korkas, Michalis Terzopoulos, Christos Tsaknakis, Aliki Stefanopoulou, Iakovos Michailidis, and Elias Kosmatopoulos. Chargym: An ev charging station model for controller benchmarking. In IFIP International Conference on Artificial Intelligence App...

    2022

  9. [17]

    Equinox: neural networks in JAX via callable PyTrees and filtered transformations

    Patrick Kidger and Cristian Garcia. Equinox: neural networks in JAX via callable PyTrees and filtered transformations. Differentiable Programming workshop at Neural Information Processing Systems 2021, 2021

    2021

  10. [18]

    Pgx: Hardware-accelerated parallel game simulators for reinforcement learning

    Sotetsu Koyamada, Shinri Okano, Soichiro Nishimori, Yu Murata, Keigo Habara, Haruka Kita, and Shin Ishii. Pgx: Hardware-accelerated parallel game simulators for reinforcement learning. In Advances in Neural Information Processing Systems, volume 36, pp.\ 45716--45743, 2023

    2023

  11. [19]

    gymnax : A JAX -based reinforcement learning environment library, 2022

    Robert Tjarko Lange. gymnax : A JAX -based reinforcement learning environment library, 2022. URL http://github.com/RobertTLange/gymnax

    2022

  12. [20]

    An Analysis of Battery Degradation in the Integrated Energy Storage System with Solar Photovoltaic Generation

    Munsu Lee, Jinhyeong Park, Sun-Ik Na, Hyung Sik Choi, Byeong-Sik Bu, and Jonghoon Kim. An Analysis of Battery Degradation in the Integrated Energy Storage System with Solar Photovoltaic Generation . Electronics, 9 0 (4): 0 701, April 2020 a . ISSN 2079-9292. doi:10.3390/electr...

    2020 doi

  13. [21]

    Lee, Sunash Sharma, Daniel Johansson, and Steven H

    Zachary J. Lee, Sunash Sharma, Daniel Johansson, and Steven H. Low. ACN-Sim : An Open-Source Simulator for Data-Driven Electric Vehicle Charging Research . https://arxiv.org/abs/2012.02809v2, December 2020 b

    2012 arXiv

  14. [22]

    Lee, George Lee, Ted Lee, Cheng Jin, Rand Lee, Zhi Low, Daniel Chang, Christine Ortega, and Steven H

    Zachary J. Lee, George Lee, Ted Lee, Cheng Jin, Rand Lee, Zhi Low, Daniel Chang, Christine Ortega, and Steven H. Low. Adaptive Charging Networks : A Framework for Smart Electric Vehicle Charging . IEEE Transactions on Smart Grid, 12 0 (5): 0 4339--4350, September 2021. ISSN 19...

    2021

  15. [23]

    Yang Li, Meng Han, Zhen Yang, and Guoqing Li. Coordinating Flexible Demand Response and Renewable Uncertainties for Scheduling of Community Integrated Energy Systems With an Electric Vehicle Charging Station : A Bi-Level Approach . IEEE Transactions on Sustainable Energy, 12 0...

    2021

  16. [24]

    luchris429/purejaxrl, September 2024

    Chris Lu. luchris429/purejaxrl, September 2024. URL https://github.com/luchris429/purejaxrl. original-date: 2023-02-25T15:38:11Z

    2024

  17. [25]

    Discovered policy optimisation

    Chris Lu, Jakub Kuba, Alistair Letcher, Luke Metz, Christian Schroeder de Witt, and Jakob Foerster. Discovered policy optimisation. Advances in Neural Information Processing Systems, 35: 0 16455--16468, 2022

    2022

  18. [26]

    A two-step approach to optimal management of electric vehicle parking lots

    Mohammad Javad Mirzaei and Ahad Kazemi. A two-step approach to optimal management of electric vehicle parking lots. Sustainable Energy Technologies and Assessments, 46: 0 101258, August 2021. ISSN 22131388. doi:10.1016/j.seta.2021.101258

    2021

  19. [27]

    XL and-minigrid: Scalable meta-reinforcement learning environments in JAX

    Alexander Nikulin, Vladislav Kurenkov, Ilya Zisman, Viacheslav Sinii, Artem Agarkov, and Sergey Kolesnikov. XL and-minigrid: Scalable meta-reinforcement learning environments in JAX . In Intrinsically-Motivated and Open-Ended Learning Workshop, NeurIPS2023, 2023. URL https://o...

    2023

  20. [28]

    Y lmaz, Peter Palensky, and Pedro P

    Stavros Orfanoudakis, Cesar Diaz-Londono , Yunus E. Y lmaz, Peter Palensky, and Pedro P. Vergara. EV2Gym : A Flexible V2G Simulator for EV Smart Charging Research and Benchmarking , April 2024

    2024

  21. [29]

    Navix: Scaling minigrid environments with jax

    Eduardo Pignatelli, Jarek Liesen, Robert Tjarko Lange, Chris Lu, Pablo Samuel Castro, and Laura Toni. Navix: Scaling minigrid environments with jax. arXiv preprint arXiv:2407.19396, 2024

    2024 arXiv

  22. [30]

    Reinforcement learning for sustainable energy: A survey

    Koen Ponse, Felix Kleuker, M \'a rton Fej \'e r, \'A lvaro Serra-G \'o mez, Aske Plaat, and Thomas Moerland. Reinforcement learning for sustainable energy: A survey. arXiv preprint arXiv:2407.18597, 2024

    2024 arXiv

  23. [31]

    Stable- Baselines3 : Reliable Reinforcement Learning Implementations

    Antonin Raffin, Ashley Hill, Adam Gleave, Anssi Kanervisto, Maximilian Ernestus, and Noah Dormann. Stable- Baselines3 : Reliable Reinforcement Learning Implementations . Journal of Machine Learning Research, 22 0 (268): 0 1--8, 2021. ISSN 1533-7928. URL http://jmlr.org/papers/...

    2021

  24. [32]

    Grid-integrated solutions for sustainable ev charging: a comparative study of renewable energy and battery storage systems

    Anis ur Rehman, Haris M Khalid, and SM Muyeen. Grid-integrated solutions for sustainable ev charging: a comparative study of renewable energy and battery storage systems. Frontiers in Energy Research, 12: 0 1403883, 2024

    2024

  25. [33]

    Evlibsim: A tool for the simulation of electric vehicles’ charging stations using the evlib library

    Emmanouil S Rigas, Sotiris Karapostolakis, Nick Bassiliades, and Sarvapali D Ramchurn. Evlibsim: A tool for the simulation of electric vehicles’ charging stations using the evlib library. Simulation Modelling Practice and Theory, 87: 0 99--119, 2018

    2018

  26. [34]

    Jaxmarl: Multi-agent rl environments in jax

    Alexander Rutherford, Benjamin Ellis, Matteo Gallici, Jonathan Cook, Andrei Lupu, Gardar Ingvarsson, Timon Willi, Akbir Khan, Christian Schroeder de Witt, Alexandra Souly, Saptarashmi Bandyopadhyay, Mikayel Samvelyan, Minqi Jiang, Robert Tjarko Lange, Shimon Whiteson, Bruno La...

    2023 arXiv

  27. [35]

    S. Saxena. Vehicle-to-grid simulator, version 00, November 2013. [Online]. Available: https://www.osti.gov/biblio/1437011

    2013

  28. [36]

    Imagenet moment for reinforcement learning?, feb 2025

    Tim Scarfe, Jakob Foerster, and Chris Lu. Imagenet moment for reinforcement learning?, feb 2025. URL https://www.dropbox.com/scl/fi/yqjszhntfr00bhjh6t565/JAKOB.pdf?rlkey=scvny4bnwj8th42fjv8zsfu2y&e=1&dl=0. Machine Learning Streettalk podcast episode

    2025

  29. [37]

    Avoiding fusion plasma tearing instability with deep reinforcement learning

    Jaemin Seo, SangKyeun Kim, Azarakhsh Jalalvand, Rory Conlin, Andrew Rothstein, Joseph Abbate, Keith Erickson, Josiah Wai, Ricardo Shousha, and Egemen Kolemen. Avoiding fusion plasma tearing instability with deep reinforcement learning. Nature, 626 0 (8000): 0 746--751, 2024

    2024

  30. [38]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto. Reinforcement Learning: An Introduction. Adaptive Computation and Machine Learning Series. The MIT Press, Cambridge, Massachusetts, second edition edition, 2018. ISBN 978-0-262-03924-6

    2018

  31. [39]

    Grid and user-optimized planning of charging processes of an electric vehicle fleet using a quantitative optimization model

    Fynn Welzel, Carl-Friedrich Klinck, Yannick Pohlmann, and Mats Bednarczyk. Grid and user-optimized planning of charging processes of an electric vehicle fleet using a quantitative optimization model. Applied Energy, 290: 0 116717, May 2021. ISSN 03062619. doi:10.1016/j.apenerg...

    2021

  32. [40]

    Improving Sample Efficiency in Model - Free Reinforcement Learning from Images , July 2020

    Denis Yarats, Amy Zhang, Ilya Kostrikov, Brandon Amos, Joelle Pineau, and Rob Fergus. Improving Sample Efficiency in Model - Free Reinforcement Learning from Images , July 2020. URL http://arxiv.org/abs/1910.01741. arXiv:1910.01741 [cs]

    2020 arXiv

  33. [41]

    Learning to Operate an Electric Vehicle Charging Station Considering Vehicle-Grid Integration

    Zuzhao Ye, Yuanqi Gao, and Nanpeng Yu. Learning to Operate an Electric Vehicle Charging Station Considering Vehicle-Grid Integration . IEEE Transactions on Smart Grid, 13 0 (4): 0 3038--3048, July 2022. ISSN 1949-3061. doi:10.1109/TSG.2022.3165479

    2022

  34. [42]

    Sustaingym: Reinforcement learning environments for sustainable energy systems

    Christopher Yeh, Victor Li, Rajeev Datta, Julio Arroyo, Nicolas Christianson, Chi Zhang, Yize Chen, Mohammad Mehdi Hosseini, Azarang Golmohammadi, Yuanyuan Shi, et al. Sustaingym: Reinforcement learning environments for sustainable energy systems. Advances in Neural Informatio...

    2024

  35. [43]

    Optimal Charging Scheduling by Pricing for EV Charging Station With Dual Charging Modes

    Yongmin Zhang, Pengcheng You, and Lin Cai. Optimal Charging Scheduling by Pricing for EV Charging Station With Dual Charging Modes . IEEE Transactions on Intelligent Transportation Systems, 20 0 (9): 0 3386--3396, September 2019. ISSN 1558-0016. doi:10.1109/TITS.2018.2876287

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.