Application of LLM Guided Reinforcement Learning in Formation Control with Collision Avoidance

Chenhao Yao; Chi Zhu; Xiaoxu Liu; Zike Yuan

REVIEW 5 major objections 5 minor 25 references

Application of LLM Guided Reinforcement Learning in Formation Control with Collision Avoidance

T0 review · 5 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 An LLM that writes and rewrites reward functions drives a multi-agent formation-control policy to 100% success.

desk verdict Sensible incremental extension of LLM-generated rewards to multi-agent formation control, but the headline comparison is confounded by an adaptive refinement loop that the human baseline does not get. read the letter →

arxiv 2507.16382 v1 pith:ZHZSNRPP submitted 2025-07-22 cs.RO cs.AI

Chenhao Yao , Zike Yuan , Xiaoxu Liu , Chi Zhu This is my paper · ORCID

classification cs.ROcs.AI

keywords LLM-guidedrewarddesignmulti-agentreinforcementlearningformationcontrolcollisionavoidanceshapingPPOsim-to-realtransfertask-levelevaluationmetrics

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

This paper argues that the hardest part of multi-agent reinforcement learning for formation control with collision avoidance—designing a reward function that balances several conflicting objectives—can be handed to a large language model. The LLM writes an initial reward from a plain-language task description and the agents' observation structure, and then, after each training round, it receives five task-level evaluation metrics (success rate, hazard incidents, formation error, total time, average acceleration) and rewrites the reward accordingly. In three refinement iterations the success rate rises from 60% to 100% in a dense obstacle environment, and the final policy outperforms both a classical reciprocal-velocity-obstacle baseline and a human-designed reward in success rate, average time, and formation error. The authors also show the learned policy transfers to a Gazebo simulation and to physical robots.

What carries the argument

The carrying mechanism is a closed-loop reward-design loop. The LLM is given a prompt containing the agent observation structure (own state, obstacle states, and communication-derived formation information encoded through the symmetric normalized Laplacian of the agent graph), a task list with priorities, and tips; it outputs an executable reward function. After PPO training with GAE, the policy is evaluated on five objective-level metrics, and those metrics—not the reward—are appended to the prompt so the LLM rewrites the reward. This design is intended to avoid the local-optima trap of feeding reward magnitudes back, which the authors attribute to prior one-shot LLM reward generators.

What would settle it

Train the human-designed reward function from [10] under the same protocol as Algorithm 2—same metrics, same number of rewrites, same training budget—and compare final success rates. If the human-tuned reward also reaches 100% success with comparable formation error, the LLM's specific role is not established. Alternatively, freeze the LLM's prompt after the first rewrite and check whether performance stops improving.

Watch

Extended reading notes

Core claim

The paper's central claim is that LLM-guided online reward refinement, driven by high-level task metrics rather than raw reward magnitudes, lets a multi-agent PPO policy solve formation control with collision avoidance in dynamic environments more efficiently than reward functions written once by a human. The authors demonstrate this with three iterations of the loop in which the LLM produces a reward function, the policy is trained until its loss converges, five task-level metrics are evaluated, and those metrics are fed back for the next rewrite. The final iteration reaches a 100% success rate across evaluation episodes, with formation error dropping from 74.6 to 27.2, while the comparison methods ORCA-F and the human-designed baseline reach 79% and 93% success respectively. The method's practical value is supported by sim-to-sim and sim-to-real deployment on Mecanum-wheel robots.

Load-bearing premise

The claim that LLM-guided rewards outperform human-designed ones rests on comparing an adaptively rewritten reward against a static baseline; if the human reward were given the same iterative tuning on the same metrics, the reported advantage could disappear.

Editorial extensions

If this is right

If the central claim holds, multi-objective MAS tasks like formation control no longer require laborious manual reward shaping; an LLM can start from a simple objective and refine the reward on its own.
Because the LLM is used only during training, the deployed policy carries no extra inference cost, so the approach fits real-time robot control.
The reported 100% success rate, after only three rewrites, suggests the metric-based feedback loop converges faster than training with a fixed reward.
The successful sim-to-real transfer indicates the reward loop produces policies that behave sensibly on physical hardware, not just in simulation.

Reading between the lines

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

An implication the authors leave implicit: the same metric-driven LLM loop should apply to other MARL tasks with conflicting objectives (e.g., coverage, tracking, or sorting), where only the observation class and the metric list would need to change.
A fairer test of the LLM's contribution would give the human-designed reward the same adaptive retraining budget; until then, the advantage over the human baseline is partly a comparison between adaptive and static tuning.
If the success plateau at iteration 3 holds across random seeds, a natural next experiment is to test whether the LLM's refinements generalize to unseen obstacle configurations, which the current evaluation does not report.

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.

Desk Editor's Note

read the letter

The paper applies an established idea — LLM-generated reward functions (Eureka, Text2Reward) — to a new setting: multi-agent formation control with collision avoidance. The genuinely new bit is that they refine the reward online using task-level evaluation metrics (success rate, formation error, hazard incidents) rather than raw reward magnitude, which is a sensible fix for a known failure mode in Eureka-style feedback. The method is clearly described, and the authors are upfront about limitations in the conclusion.

The soft spot is the comparison. In Algorithm 2, the LLM rewrites the reward for up to three iterations based on exactly the metrics that Table II reports. The human-designed baseline is trained once with a static reward. So Table II's advantage (95% vs 93% success, 11.5 vs 14.5s, 26.7 vs 37.4 formation error) could just as easily come from the extra tuning budget as from the LLM's reward-generation skill. The stress-test note has this right; the paper does not address it. The 100% success in Table I is a stopping condition, not a steady-state estimate, and it is averaged over only 20 episodes. Table II reports no variance or significance tests. The baseline reward is never described, and the ORCA-F citation (ref [24]) points to a deep-RL collision avoidance paper, not ORCA-F. Real-world validation is qualitative, with only snapshots.

None of this makes the method useless. If the authors release the prompts and code, describe the baseline reward, and run a fair comparison where the human-designed reward gets the same iterative tuning budget (or a simple automatic weight search), the result would be a useful engineering contribution. The authors are honest about the limitations of LLM-based reward design losing track of earlier objectives, which is more than many papers in this space do. The novelty is modest but real.

Who this is for: robotics practitioners working on MAS reward design and people building on Eureka-style workflows. It does not change theory, but it is a reasonable data point. I would cite it if the comparison is fixed. I would send it to peer review — it deserves referee time, but the central claim needs substantial revision. The paper shows clear thinking and honest engagement with prior work; the issues are experimental rigor, not incoherence.

Referee Report

5 major / 5 minor

Summary. The paper proposes a framework for multi-agent formation control with collision avoidance (FCCA) in which a large language model (LLM, specifically Qwen2.5-72B) generates the initial reward function for a multi-agent PPO policy and then iteratively rewrites that reward based on high-level task metrics (success rate, hazard incidents, formation error, total time, average acceleration) rather than on raw reward magnitudes. The authors claim that after three refinement iterations the method reaches 100% success in simulation, outperforms a human-designed reward baseline and ORCA-F in success rate, average time, and formation error (Table II), and transfers to Gazebo simulation and real-world robots. The paper also includes a qualitative real-world deployment with Mecanum-wheel robots using OptiTrack localization and Jetson AGX Orin compute.

Significance. If the stated results held, this would be a useful incremental contribution to automated reward design for multi-agent RL: the idea of closing the reward-tuning loop with task-level metrics is sensible and has practical appeal, and the paper demonstrates a plausible pipeline from LLM-generated reward to trained deployment. The main strengths are the concrete algorithm descriptions and the inclusion of both simulation and physical-robot experiments. However, the empirical support for the central comparative claim is currently weak: the LLM method is given an iterative refinement budget while the human-designed baseline is static, no variance or significance measures are reported, the ORCA-F baseline is mis-cited, and the real-world validation is qualitative. These issues are fixable within the scope of the manuscript, which is why I recommend major revision rather than rejection.

major comments (5)

[Section V.A, Table II, Algorithm 2] The central comparative claim that the LLM-guided reward outperforms human-designed rewards is confounded by an unequal adaptation budget. The LLM method runs Algorithm 2, in which the reward function is rewritten multiple times based on the very evaluation metrics (success rate, formation error, total time) that are then reported in Table II, while the human-designed baseline [10] is trained with a static reward. The observed differences (95% vs 93% success, 11.5 s vs 14.5 s, 26.7 vs 37.4 formation error) may therefore reflect the iterative tuning protocol rather than the LLM's reward-generation ability. Please add a matched baseline in which the human-designed reward (or an automated weight-search baseline) receives the same iterative refinement budget on the same metrics, or report the best single-shot LLM reward without online tuning.
[Section V.A, Tables I and II] No measure of variance is reported for any of the headline results. Table I reports averages over only 20 evaluation episodes per iteration, and Table II reports averages over 3 seeds x 300 episodes without standard deviations or significance tests. With 900 episodes, the 2-point success-rate difference between LLM (95%) and human-designed (93%) is not statistically significant (a two-proportion z-test gives p ≈ 0.07), and the 100% figure in Table I is a stopping condition, not a steady-state estimate. Please report per-seed results, standard errors, and significance tests, and evaluate the final policy on a held-out set of obstacle configurations.
[Section V.A, Table II, reference [24]] The baseline labeled ORCA-F is cited to [24], which is Chen et al., 'Decentralized non-communicating multiagent collision avoidance with deep reinforcement learning' (ICRA 2017), not an ORCA-family method. This citation mismatch makes it impossible to verify the baseline implementation. Please provide the correct reference for ORCA-F (e.g., the original ORCA paper or an ORCA-based formation controller), and describe the baseline's parameters and how it was adapted to the formation-control task.
[Section V.B] The real-world validation is entirely qualitative. Figure 7 shows snapshots, but the paper reports no quantitative measures for the physical experiments, such as number of trials, success rate, minimum distance to obstacles, formation error over time, or completion time. The claim that the approach is 'validated in real-world settings' is therefore unsupported. Please report repeated real-world runs with the same metrics used in simulation.
[Section IV.C, Algorithm 2] The evaluation metrics used for reporting are fed back into the reward-refinement loop (Algorithm 2 lines 3-5), so the reported success rate and formation error are fitted quantities, not independent estimates. The LLM is explicitly instructed to adjust the reward to improve these metrics (Fig. 4), which makes the later-iteration numbers a form of test-set optimization. At minimum, the final evaluation should be performed on a separate held-out set of environments and obstacle trajectories that were never used in the feedback loop.

minor comments (5)

[Abstract, Section I] The abstract contains a grammatical fragment ('However, when confronted with the complex objective of Formation Control with Collision Avoidance (FCCA): designing an effective reward function ...') and should be rewritten for clarity.
[Section III.B, Eq. (5)] The TD-error definition has an index error: δ_t = r_t + γV(s_{t+1}) − V(s_t) is standard; the printed V(s_{t−1}) appears to be a typo. Also, the symbol r_t is used both for the reward at time t and the probability ratio in Eq. (8); please disambiguate.
[Algorithm 2] The input 'policy summary' used to update the prompt (line 4) is never defined; please specify what information it contains and how it is generated.
[Section V.A, Fig. 5] The reward-curve comparison is hard to interpret because the LLM reward function changes at each iteration, and rewards from different reward functions are not on a common scale; please plot a task-level metric or normalize the curves.
[Section IV.B, Algorithm 1] The sentence order implies secondary objectives are only considered after primary ones are achieved, but Algorithm 1 uses a convergence threshold η that is never given a numerical value; please specify η and the termination criterion for each iteration.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported success/time/formation-error gains are fitted quantities: Algorithm 2 feeds exactly those metrics back to rewrite the reward, while the human-designed baseline is static.

fitted input called prediction [Section IV.C (Algorithm 2), Section V.A (Table I, Table II, Fig. 4)]
"Evaluate task metrics Mk (e.g., formation error, hazard, time); Update prompt Pk with Mk and policy summary; Generate new reward Rk ← LLM(Pk); ... Each iteration consists of 20 episodes, with the reported results being the average performance across these episodes. These metrics will assist in refining the reward function, ensuring it more accurately reflects the desired outcomes and improves the performance of the agents."

The reported outcomes—success rate, average time, and formation error—are exactly the evaluation metrics Mk that Algorithm 2 feeds back to the LLM to rewrite the reward. Fig. 4 shows the LLM making targeted adjustments ('Significantly increased the penalty for collisions', 'We increased the penalty for formation errors') in response to those metrics. Thus the numbers in Table I and the LLM row of Table II are values of the objective function being optimized by the reward-rewriting loop, not independent predictions of the method's quality. The 100% success rate at iteration 3 is also the stopping threshold in Algorithm 1 (if s_k >= eta, break), so it is a stopping condition rather than a stable estimate.

full rationale

The formal RL derivation (MDP, PPO, GAE) is standard and self-contained; there is no problematic self-citation or imported uniqueness theorem. The circularity is concentrated in the empirical evaluation protocol: the metrics that drive reward rewriting in Algorithm 2 are identical to the metrics reported in Tables I and II. The LLM's reward is iteratively adjusted until success rate reaches the threshold and formation/time metrics improve, so those figures are fitted quantities rather than held-out predictions. The human-designed baseline in Table II is not given the same iterative tuning budget, making the reported superiority an apples-to-oranges comparison. The 100% success rate is a stopping condition, not a steady-state estimate, and the 20-episode feedback loop makes the final numbers especially susceptible to this confound. Consequently, the central comparative claim is partially circular, even though the underlying idea—LLM generates rewards and uses task-level metrics for online adjustment—is not empty. A fair comparison would give both methods the same number of reward revisions or evaluate the final reward in a held-out setting with significance testing. The real-world deployment is a useful sanity check but does not resolve the confounded comparison.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the LLM producing valid reward code from text prompts, a stochastic and unverified process, and on evaluation metrics that are also used as the optimization target. These are not standard math axioms, but domain assumptions about the reliability of LLM outputs and the validity of the metrics as proxies.

free parameters (4)

Reward component weights = not disclosed; adjusted by LLM across iterations
Each iteration's reward function is a weighted sum of terms (goal, collision, formation, velocity, time); the LLM adjusts weights based on evaluation metrics, effectively fitting these weights to the observed success rate and formation error (Section IV.C, Fig. 4).
Success threshold eta = 100%
Algorithm 1 stops iterating when success rate reaches eta; in the experiments, iterations continue until 100% is reached, so eta functions as a target rather than a measured outcome (Algorithm 1, Section V.A).
Number of refinement iterations = 4 (iterations 0 to 3)
The loop was run until 100% success appeared, a stopping-time choice rather than a fixed budget (Algorithm 2, Section V.A).
Number of evaluation episodes per iteration = 20
Success rates in Table I are averages over 20 episodes; this small sample inflates the precision of the reported 100% (Section V.A).

assumptions (5)

standard math The MDP/PPO/GAE formulation (Eqs. 5-8) is assumed to be correct and to converge under the given hyperparameters.
The paper uses standard PPO with GAE, and does not prove convergence; it relies on standard RL theory (Section III.B, Section IV.C).
domain assumption The formation error in Eq. (4), based on symmetric normalized Laplacian distance, is a valid proxy for formation quality under dynamic scaling and rotation.
The metric is borrowed from [20] and is used to evaluate formation maintenance, but its validity in incomplete or changing communication graphs is not demonstrated (Section IV.A).
ad hoc to paper The LLM can reliably convert textual task descriptions and numeric evaluation feedback into executable reward code without human intervention.
The method depends on the stochastic outputs of Qwen2.5-72B; no guarantees or robustness analysis are provided (Section IV.B, Fig. 4).
ad hoc to paper The high-level evaluation metrics (success rate, hazard incidents, formation error, total time, average acceleration) are sufficient to capture task success and guide reward improvement.
These metrics are chosen by the authors and used both as feedback and as reported outcomes; their sufficiency is assumed (Section IV.C, Algorithm 2).
domain assumption The simulation environment (Gazebo) is faithful enough that policies trained in simulation transfer to the real-world arena without additional adaptation.
Real-world deployment is shown qualitatively; no quantitative sim-to-real gap measurement is provided (Section V.B).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Application of LLM Guided Reinforcement Learning in Formation Control with Collision Avoidance." pith.science (2026). https://pith.science/paper/ZHZSNRPP

@misc{pith2026250716382,
  author       = {Pith},
  title        = {Pith review of: Application of LLM Guided Reinforcement Learning in Formation Control with Collision Avoidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZHZSNRPP}},
  note         = {Machine review of arXiv:2507.16382}
}

read the original abstract

Multi-Agent Systems (MAS) excel at accomplishing complex objectives through the collaborative efforts of individual agents. Among the methodologies employed in MAS, Multi-Agent Reinforcement Learning (MARL) stands out as one of the most efficacious algorithms. However, when confronted with the complex objective of Formation Control with Collision Avoidance (FCCA): designing an effective reward function that facilitates swift convergence of the policy network to an optimal solution. In this paper, we introduce a novel framework that aims to overcome this challenge. By giving large language models (LLMs) on the prioritization of tasks and the observable information available to each agent, our framework generates reward functions that can be dynamically adjusted online based on evaluation outcomes by employing more advanced evaluation metrics rather than the rewards themselves. This mechanism enables the MAS to simultaneously achieve formation control and obstacle avoidance in dynamic environments with enhanced efficiency, requiring fewer iterations to reach superior performance levels. Our empirical studies, conducted in both simulation and real-world settings, validate the practicality and effectiveness of our proposed approach.

Figures

Figures reproduced from arXiv: 2507.16382 by the authors.

**Figure 1.** Overview of our method Initially, MARL was primarily applied in video games, like StarCraft II [6] and DOTA II [7]. Furthermore, in [8] , the researchers conducted a series of extensive experiments utilizing the Proximal Policy Optimization (PPO) algorithm across a multitude of distinct cooperative multi-agent benchmarks. The experimental results consistently demonstrated superior performance, with the level of coo… view at source ↗

**Figure 2.** Diagram of the agents observatios typically gathers its own local observations, observations of obstacles within its vicinity, as well as data acquired through communication with neighboring agents, a diagram is shown in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Training framework by using Actor-Critic method [22] and PPO [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: summarizes the conversation content with the LLM. The initial reward function generated in the 0th iteration could only accomplish the tasks of reaching the destination and avoiding obstacles, without considering the interactions among various reward components. Starti…

**Figure 5.** Figure 5: LLM-Generated vs Human-Designed Reward Functions [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Snapshots of simulation [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: Snapshots of real-world deployment [9] Z. Sui, Z. Pu, J. Yi, and S. Wu, “Formation control with collision avoidance through deep reinforcement learning using model-guided demonstration,” IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 6, pp. 235…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 18 canonical work pages

[24]

Decentralized non- communicating multiagent collision avoidance with deep reinforce- ment learning,

Y . F. Chen, M. Liu, M. Everett, and J. P. How, “Decentralized non- communicating multiagent collision avoidance with deep reinforce- ment learning,” in 2017 IEEE International Conference on Robotics and Automation (ICRA) , pp. 285–292, 2017

work page 2017
[10]

Relative distributed formation and obstacle avoidance with multi- agent reinforcement learning,

Y . Yan, X. Li, X. Qiu, J. Qiu, J. Wang, Y . Wang, and Y . Shen, “Relative distributed formation and obstacle avoidance with multi- agent reinforcement learning,” in 2022 International Conference on Robotics and Automation (ICRA) , pp. 1661–1667, 2022

work page 2022
[1]

Reciprocal col- lision avoidance with acceleration-velocity obstacles,

J. van den Berg, J. Snape, S. J. Guy, and D. Manocha, “Reciprocal col- lision avoidance with acceleration-velocity obstacles,” in 2011 IEEE International Conference on Robotics and Automation, pp. 3475–3482, 2011

work page 2011
[2]

An improved artificial potential field method for path planning and formation control of the multi-uav systems,

Z. Pan, C. Zhang, Y . Xia, H. Xiong, and X. Shao, “An improved artificial potential field method for path planning and formation control of the multi-uav systems,” IEEE Transactions on Circuits and Systems II: Express Briefs , vol. 69, no. 3, pp. 1129–1133, 2022

work page 2022
[3]

Proximal policy optimization algorithms,

J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” arXiv preprint arXiv:1707.06347, 2017

arXiv 2017
[4]

Impala: Scalable distributed deep-rl with importance weighted actor-learner architec- tures,

L. Espeholt, H. Soyer, R. Munos, K. Simonyan, V . Mnih, T. Ward, Y . Doron, V . Firoiu, T. Harley, I. Dunning et al. , “Impala: Scalable distributed deep-rl with importance weighted actor-learner architec- tures,” in International conference on machine learning , pp. 1407–

work page
[5]

Monotonic value function factorisation for deep multi- agent reinforcement learning,

T. Rashid, M. Samvelyan, C. S. De Witt, G. Farquhar, J. Foerster, and S. Whiteson, “Monotonic value function factorisation for deep multi- agent reinforcement learning,” Journal of Machine Learning Research, vol. 21, no. 178, pp. 1–51, 2020

work page 2020
[6]

Grandmaster level in starcraft ii using multi-agent reinforcement learning,

O. Vinyals, I. Babuschkin, W. M. Czarnecki, M. Mathieu, A. Dudzik, J. Chung, D. H. Choi, R. Powell, T. Ewalds, P. Georgiev et al. , “Grandmaster level in starcraft ii using multi-agent reinforcement learning,” Nature, vol. 575, no. 7782, pp. 350–354, 2019

work page 2019

Show all 25 references

[7]

Dota 2 with large scale deep reinforcement learning,

C. Berner, G. Brockman, B. Chan, V . Cheung, P. D˛ ebiak, C. Dennison, D. Farhi, Q. Fischer, S. Hashme, C. Hesse et al., “Dota 2 with large scale deep reinforcement learning,” arXiv preprint arXiv:1912.06680 , 2019

1912 arXiv
[8]

The surprising effectiveness of ppo in cooperative multi- agent games,

C. Yu, A. Velu, E. Vinitsky, J. Gao, Y . Wang, A. Bayen, and Y . WU, “The surprising effectiveness of ppo in cooperative multi- agent games,” in Advances in Neural Information Processing Systems , S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, Eds., vol. 35...

2022
[9]

Formation control with collision avoidance through deep reinforcement learning using model-guided demonstration,

Z. Sui, Z. Pu, J. Yi, and S. Wu, “Formation control with collision avoidance through deep reinforcement learning using model-guided demonstration,” IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 6, pp. 2358–2372, 2021

2021
[11]

Do as i can, not as i say: Grounding language in robotic affordances,

M. Ahn, A. Brohan, N. Brown, Y . Chebotar, O. Cortes, B. David, C. Finn, C. Fu, K. Gopalakrishnan, K. Hausman et al. , “Do as i can, not as i say: Grounding language in robotic affordances,” arXiv preprint arXiv:2204.01691, 2022

2022 arXiv
[12]

Executable code actions elicit better LLM agents,

X. Wang, Y . Chen, L. Yuan, Y . Zhang, Y . Li, H. Peng, and H. Ji, “Executable code actions elicit better LLM agents,” in Forty-first International Conference on Machine Learning , 2024

2024
[13]

Boosting efficient reinforcement learning for vision-and-language navigation with open- sourced llm,

J. Wang, T. Wang, W. Cai, L. Xu, and C. Sun, “Boosting efficient reinforcement learning for vision-and-language navigation with open- sourced llm,” IEEE Robotics and Automation Letters , vol. 10, no. 1, pp. 612–619, 2025

2025
[14]

Text2Reward: Reward Shaping with Language Models for Reinforcement Learning,

T. Xie, S. Zhao, C. H. Wu, Y . Liu, Q. Luo, V . Zhong, Y . Yang, and T. Yu, “Text2Reward: Reward Shaping with Language Models for Reinforcement Learning,” arXiv e-prints, 2023

2023
[15]

Eureka: Human- level reward design via coding large language models,

Y . J. Ma, W. Liang, G. Wang, D.-A. Huang, O. Bastani, D. Ja- yaraman, Y . Zhu, L. Fan, and A. Anandkumar, “Eureka: Human- level reward design via coding large language models,” arXiv preprint arXiv:2310.12931, 2023

2023 arXiv
[16]

Adaptive finite-time tracking control of nonholonomic multirobot formation systems with limited field-of- view sensors,

S.-L. Dai, K. Lu, and J. Fu, “Adaptive finite-time tracking control of nonholonomic multirobot formation systems with limited field-of- view sensors,” IEEE Transactions on Cybernetics, vol. 52, no. 10, pp. 10 695–10 708, Oct. 2022

2022
[17]

Racer: Rapid collaborative explo- ration with a decentralized multi-uav system,

B. Zhou, H. Xu, and S. Shen, “Racer: Rapid collaborative explo- ration with a decentralized multi-uav system,” IEEE Transactions on Robotics, vol. 39, no. 3, pp. 1816–1835, 2023

2023
[18]

An overview of recent advances in coordinated control of multiple autonomous surface vehicles,

Z. Peng, J. Wang, D. Wang, and Q.-L. Han, “An overview of recent advances in coordinated control of multiple autonomous surface vehicles,” IEEE Transactions on Industrial Informatics, vol. 17, no. 2, pp. 732–745, Feb. 2021

2021
[19]

M. L. Puterman, Markov decision processes: discrete stochastic dy- namic programming. John Wiley & Sons, 2014

2014
[20]

Distributed swarm trajectory optimization for formation flight in dense environments,

L. Quan, L. Yin, C. Xu, and F. Gao, “Distributed swarm trajectory optimization for formation flight in dense environments,” in 2022 International Conference on Robotics and Automation (ICRA) , pp. 4979–4985, 2022

2022
[21]

High-dimensional continuous control using generalized advantage estimation,

S. L. M. J. P. A. John Schulman, Philipp Moritz, “High-dimensional continuous control using generalized advantage estimation,” arXiv preprint arXiv:1506.02438, 2015

2015 arXiv
[22]

Asynchronous Methods for Deep Reinforcement Learning,

V . Mnih, A. Puigdomènech Badia, M. Mirza, A. Graves, T. P. Lillicrap, T. Harley, D. Silver, and K. Kavukcuoglu, “Asynchronous Methods for Deep Reinforcement Learning,” arXiv e-prints, 2016

2016
[23]

Qwen2.5 Technical Report,

A. Yang et al., “Qwen2.5 Technical Report,” arXiv e-prints, 2024

2024
[25]

Dreureka: Language model guided sim-to-real transfer,

Y . J. Ma, W. Liang, H.-J. Wang, S. Wang, Y . Zhu, L. Fan, O. Bastani, and D. Jayaraman, “Dreureka: Language model guided sim-to-real transfer,” arXiv preprint arXiv:2406.01967 , 2024

2024 arXiv

Pith tools

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

[1] [24]

Decentralized non- communicating multiagent collision avoidance with deep reinforce- ment learning,

Y . F. Chen, M. Liu, M. Everett, and J. P. How, “Decentralized non- communicating multiagent collision avoidance with deep reinforce- ment learning,” in 2017 IEEE International Conference on Robotics and Automation (ICRA) , pp. 285–292, 2017

work page 2017

[2] [10]

Relative distributed formation and obstacle avoidance with multi- agent reinforcement learning,

Y . Yan, X. Li, X. Qiu, J. Qiu, J. Wang, Y . Wang, and Y . Shen, “Relative distributed formation and obstacle avoidance with multi- agent reinforcement learning,” in 2022 International Conference on Robotics and Automation (ICRA) , pp. 1661–1667, 2022

work page 2022

[3] [1]

Reciprocal col- lision avoidance with acceleration-velocity obstacles,

J. van den Berg, J. Snape, S. J. Guy, and D. Manocha, “Reciprocal col- lision avoidance with acceleration-velocity obstacles,” in 2011 IEEE International Conference on Robotics and Automation, pp. 3475–3482, 2011

work page 2011

[4] [2]

An improved artificial potential field method for path planning and formation control of the multi-uav systems,

Z. Pan, C. Zhang, Y . Xia, H. Xiong, and X. Shao, “An improved artificial potential field method for path planning and formation control of the multi-uav systems,” IEEE Transactions on Circuits and Systems II: Express Briefs , vol. 69, no. 3, pp. 1129–1133, 2022

work page 2022

[5] [3]

Proximal policy optimization algorithms,

J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” arXiv preprint arXiv:1707.06347, 2017

arXiv 2017

[6] [4]

Impala: Scalable distributed deep-rl with importance weighted actor-learner architec- tures,

L. Espeholt, H. Soyer, R. Munos, K. Simonyan, V . Mnih, T. Ward, Y . Doron, V . Firoiu, T. Harley, I. Dunning et al. , “Impala: Scalable distributed deep-rl with importance weighted actor-learner architec- tures,” in International conference on machine learning , pp. 1407–

work page

[7] [5]

Monotonic value function factorisation for deep multi- agent reinforcement learning,

T. Rashid, M. Samvelyan, C. S. De Witt, G. Farquhar, J. Foerster, and S. Whiteson, “Monotonic value function factorisation for deep multi- agent reinforcement learning,” Journal of Machine Learning Research, vol. 21, no. 178, pp. 1–51, 2020

work page 2020

[8] [6]

Grandmaster level in starcraft ii using multi-agent reinforcement learning,

O. Vinyals, I. Babuschkin, W. M. Czarnecki, M. Mathieu, A. Dudzik, J. Chung, D. H. Choi, R. Powell, T. Ewalds, P. Georgiev et al. , “Grandmaster level in starcraft ii using multi-agent reinforcement learning,” Nature, vol. 575, no. 7782, pp. 350–354, 2019

work page 2019