Pith. sign in

REVIEW 3 major objections 5 minor 32 references

Motion Control in Multi-Rotor Aerial Robots Using Deep Reinforcement Learning

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

Pith's one-line read A reinforcement-learning controller can navigate a simulated drone to waypoints with centimeter accuracy and adapt to changing mass.

desk verdict A useful but incremental Simulink study of TD3 plus curriculum learning for UAV waypoint navigation; the central mass-adaptation claim is plausible but the observation-space ablation is not cleanly isolated. read the letter →

arxiv 2502.05996 v2 pith:4RI4U5J5 submitted 2025-02-09 cs.RO cs.AI

classification cs.ROcs.AI
keywords deepreinforcementlearningmulti-rotorUAVcontroladditivemanufacturingwaypointnavigationcurriculumtwindelayeddeterministicpolicygradientmassadaptationsimulation
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 a deep reinforcement learning agent can control a multi-rotor drone performing waypoint navigation for additive manufacturing, and that two design choices make the control robust to the mass changes that occur during material deposition. The authors compare DDPG and TD3 in a MATLAB Simulink simulation and report that TD3 is the better controller, reaching a fixed waypoint with an average positional error of 1.804e-2 m and a 96% success rate versus 5.643e-2 m and 82% for DDPG. They show that a curriculum that gradually increases task complexity raises TD3's multi-waypoint success from 66% to 87%. Their central empirical claim is that adding x/y/z accelerations to the observation space lets the agent infer mass variations through F = ma, raising success under variable mass from 26-29% to 94%. If correct, the result is a scalable simulation-to-reality path for autonomous aerial material deposition.

What carries the argument

The load-bearing object is the observation vector of equation (3), in particular the inclusion of the drone's accelerations $a_{x,t}, a_{y,t}, a_{z,t}$. Because the deposition force is known ($F = 2.135 \times 10^{-2}$ N) and the simulator's dynamics are deterministic, acceleration is a direct readout of the drone's current mass via $F = ma$; the paper claims this allows the policy to adapt its thrust commands when mass changes. The other machinery is the curriculum schedule $C(t) \in \{C_1, C_2, C_3, C_4\}$, which progressively increases task complexity and stabilizes training, and the TD3 algorithm itself, whose twin critics and delayed policy updates reduce the overestimation bias that harms DDPG.

What would settle it

Retrain the same TD3 agent with the acceleration observations replaced by realistic accelerometer signals (additive zero-mean noise with the sensor's variance and a one-step delay) and test under variable mass; if the success rate falls well below 94%, the acceleration-based mass adaptation is an artifact of the clean simulator signal. A direct flight test with a changing payload would settle the question.

Watch

Extended reading notes

Core claim

The core discovery is that a TD3 policy trained in a Simulink multirotor simulator, with roll, pitch, and thrust as actions and a reward that decays exponentially with distance to the target, can navigate waypoints with centimeter-level accuracy, and that its robustness to mass variability depends on the agent being able to observe its own accelerations. The authors model the control problem as an MDP whose state vector includes the drone's x/y/z accelerations, positional error to the target, velocities, orientation angles, and height. Curriculum learning stages the training from static waypoints to dynamic waypoints, variable mass, and external disturbances. The paper reports that without acceleration feedback the curriculum-trained agent succeeds only 26-29% of the time under variable mass, and with acceleration feedback the success rate reaches 94%, with average positional error dropping from 0.2688 m to 0.09857 m.

Load-bearing premise

The claim that acceleration feedback is what enables mass adaptation assumes the simulator gives the agent clean, near-ideal accelerations; on a real drone, accelerometer readings are noisy, delayed, and mixed with attitude dynamics, so the 94% success under variable mass would likely not transfer directly.

Editorial extensions

If this is right

  • A TD3 policy trained in simulation can serve as the control layer for waypoint navigation in drone-based additive manufacturing, reaching targets with average positional error below 2 cm in the simulator.
  • Curriculum learning is a viable way to train a single policy that handles static waypoints, dynamic waypoints, variable mass, and external disturbances without retuning.
  • Putting raw acceleration measurements into the observation space is a simple, effective way to make a learned controller aware of mass changes, at least when the measurements are clean.
  • The framework is extensible in the paper's own terms: the same MDP, reward, and curriculum can be carried toward real-world implementation, which the authors state as future work.

Reading between the lines

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

  • The acceleration-feedback result suggests a general recipe: when an unknown physical parameter (here, mass) is the only hidden variable, exposing a direct measurement of it in the observation space can convert a partially observable control problem into a nearly observable one; the same trick might apply to other variable-payload aerial tasks.
  • On real hardware, the gap between the 94% result and reality will likely be dominated by accelerometer noise and transport delay; a natural extension is to train with a realistic IMU model and measure how much accuracy degrades.
  • The paper does not isolate why TD3 beats DDPG beyond citing overestimation bias; a controlled ablation holding all else fixed would test whether the advantage is algorithmic or due to exploration noise differences.
  • The deposition-force model used, a constant reaction force of 2.135e-2 N, is small relative to a typical drone's thrust; the mass variability claim may scale only if the force and mass range reflect a real extruder, so stress testing with broader mass ranges would clarify practical limits.
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 proposes a deep reinforcement learning (DRL) framework for multirotor waypoint navigation in additive manufacturing tasks. The authors model the problem as an MDP with a state vector that includes accelerations, a three-action control space (roll, pitch, thrust), and an exponential proximity reward, and they compare DDPG with TD3 in a MATLAB Simulink UAV simulator. They report that TD3 outperforms DDPG on single-waypoint navigation (average positional error 1.804e-2 m vs. 5.643e-2 m; success 96% vs. 82%), that curriculum learning improves TD3's multi-waypoint success from 66% to 87%, and that adding x/y/z accelerations to the observation space raises TD3's success under variable mass from 26-29% to 94%. The paper concludes that TD3 with curriculum learning and acceleration feedback is a promising control approach for simulated UAV-based additive manufacturing.

Significance. If the reported results are reproducible, the paper would provide a useful empirical comparison of DDPG and TD3 for a simulated UAV control task with payload variation, and the curriculum-learning result would be a reasonable incremental contribution. The paper has concrete strengths: it reports quantitative metrics (cumulative reward, positional error, precision, success ratio), provides a hyperparameter table, uses standard off-policy algorithms, and documents the simulation setup in enough detail to be partially reconstructed. However, the most emphasized result — that acceleration feedback enables adaptation to mass variation — is not supported by a controlled ablation, and the absence of multiple training runs and statistical tests weakens the claims of consistent and significant improvement. The real-world transfer claim is also limited by the absence of process noise and the use of near-ideal acceleration signals from the simulator.

major comments (3)
  1. [§V-C and Eq. (3)] The central mass-adaptation result is not cleanly attributable to adding accelerations, because the baseline condition is not well defined. Eq. (3) already lists ax,t, ay,t, az,t as part of the proposed state vector, but §V-C states that the observation space 'was expanded to include the drone's accelerations in the x, y and z directions.' If the curriculum-trained baseline already used Eq. (3), there was no expansion; if it did not, then the baseline contradicts Eq. (3). The paper reports inconsistent pre-retraining numbers for the same condition (average cumulative reward 450 with 26% success, then 454 with 29% success), and no ablation is provided that keeps all other factors fixed while toggling only the acceleration terms. With no random seeds or error bars, the improvement from 26-29% to 94% cannot be separated from retraining, additional experience, or run-to-run variance. This issue is load-bearing because the abstract and conclusion use the 94% result to claim that TD3 'consistently' balances accuracy and success under payload changes.
  2. [§V-A and Tables II-IV] The claim that TD3's average positional error is 'significantly lower' than DDPG's is not supported by any statistical test or repeated-run evidence. Tables II, III, and IV report single values for each condition, with no number of seeds, no standard errors, and no confidence intervals. Since DRL training is stochastic, a single run cannot establish significance. Additionally, the claim that TD3 'consistently' balances stability, accuracy, and success is not supported by the presented experiments: DDPG is evaluated only on the single-waypoint task and is not tested under curriculum learning, mass variability, or multi-waypoint navigation, so no cross-condition consistency comparison between the algorithms is actually made.
  3. [§II, §III-A, and §V-C] The paper's transferability claim is not supported by the simulation model used. The text explicitly states that only measurement noise is added, with no process noise, and that noise is injected after reading the state from a deterministic multirrotor model. Thus the accelerations in Eq. (3) are an almost noiseless, delay-free proxy for mass through F = ma. On a real drone, accelerometer measurements are noisy, delayed, and coupled with attitude dynamics, so the 94% success under variable mass would likely not transfer. The paper should either state this limitation clearly as a simulation-only result or add a sensor model with realistic acceleration noise, delay, and process noise before claiming progress toward 'robust, autonomous drone control in additive manufacturing.'
minor comments (5)
  1. [§V-C, Fig. 5] The text says the success ratio rose 'from 29% to 94%,' but the same condition is earlier given as 26%, and the average cumulative reward is given as 450 and then 454; the inconsistent baseline numbers should be reconciled.
  2. [§III-A, Eq. (5)] The reward function rp = wp exp(-||p - p_desired||^2) is described as convex, but a negative exponential of a squared Euclidean norm is not convex over the whole space; this should be reworded or replaced with a precise geometric description.
  3. [§III-B, Eq. (6)] The curriculum stage C1 is defined with reward R(s,a) = -||p_target - p_current||_2, which is inconsistent with the exponential reward in Eq. (5); the authors should clarify which reward is used in the basic navigation stage.
  4. [General] There are several typographical and formatting errors, including 'randomaly' in §V-C, inconsistent spacing in 'UA Vs' and 'UAV-based', and a duplicated row in the Target Policy Smooth Model block of Table I (two 'Std. Dev.' entries).
  5. [§V-B, Table III] Table III is captioned 'Training Results' but the surrounding text and Fig. 4 describe test results; the caption should be corrected to indicate whether the numbers are from training or from the 100-trial test phase.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the paper's central results are empirical comparisons in an external simulator, and its self-citations are background only.

full rationale

The paper's core claims are empirical: TD3 outperforms DDPG in a MATLAB UAV Toolbox simulator, curriculum learning improves multi-waypoint success, and adding acceleration observations accompanies improved success under variable mass. None of these claims reduces to an input by construction: there is no fitted parameter renamed as a prediction, no uniqueness theorem imported from the authors' prior work, and no ansatz smuggled in via self-citation. The state vector in Eq. (3) includes accelerations, and Section V-C describes 'expanding' the observation space to include those same accelerations; this is textually inconsistent and the numerical comparison (450/26% vs. 454/29%) is also inconsistent. These are reproducibility and confounding concerns about whether the acceleration term is cleanly isolated, but they are not circularity: the improvement is not definitionally forced, and the self-citations [11]-[15], [19] are used for generic background about RL and UAV applications, not to justify the measured results. The DDPG/TD3 comparison and the curriculum-learning comparison are self-contained experiments against an external simulator, so the derivation chain is not circular.

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

The central empirical claims rest on the Simulink multirotor model and on several task simplifications: obstacle-free space, no yaw, no process noise, clean acceleration measurements, and a trusted attitude controller. These are reasonable for a first simulation study, but they are assumed rather than demonstrated, and they materially affect whether the reported success rates transfer to real aerial additive manufacturing. The free parameters listed above are design choices whose values are not reported, even though they influence the headline numbers.

free parameters (5)
  • reward scaling weight wp
    Appears in Eq. (5) as the multiplicative weight on the exponential distance reward; no value is reported, and cumulative reward values are sensitive to it.
  • sensor measurement noise amplitude
    A uniform noise model is mentioned in Section II, but the bounds or standard deviation are not given, despite noise being a condition that distinguishes training scenarios.
  • mass variability range
    Section V-C varies drone mass to emulate deposition, but the range of masses and how mass changes over an episode are not specified; this is the core perturbation of the main result.
  • curriculum stage transition schedule
    Curriculum levels C1-C4 are defined qualitatively in Section III-B; the conditions and timing for advancing between levels, plus differences from the no-curriculum baseline, are not specified.
  • episode termination thresholds = velocity < 0.1 m/s; height < 0.1 m; deviation > 10 m
    These thresholds define success and failure and directly affect the success ratio; they are stated but not varied.
assumptions (5)
  • domain assumption The MATLAB UAV Toolbox multirotor model with its separate attitude controller faithfully represents a real multirotor.
    Sections III-A and IV delegate all rotational dynamics to the Simulink model; the central simulation results inherit this fidelity assumption.
  • domain assumption The environment is obstacle-free and yaw control can be omitted because AM primarily needs lateral and vertical positioning.
    Section II explicitly simplifies the task this way, which limits transfer to real AM workloads that include obstacles and yaw requirements.
  • domain assumption State accelerations ax, ay, az are available and accurately reflect mass changes through F = ma.
    Section V-C relies on this to let the agent adapt thrust; real accelerometers are noisy, delayed, and coupled to attitude dynamics.
  • ad hoc to paper Only measurement noise is added to the state, and there is no process noise in the dynamics.
    Section II states that a uniform measurement noise model was used with no process noise; this modeling choice could overstate robustness to real disturbances.
  • standard math The multirotor nonlinear dynamics follow the standard modeling in Mellinger and Michael [25].
    The paper invokes this background model without rederiving it; the simulation results inherit whatever approximations that model contains.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Motion Control in Multi-Rotor Aerial Robots Using Deep Reinforcement Learning." pith.science (2026). https://pith.science/paper/4RI4U5J5

@misc{pith2026250205996,
  author       = {Pith},
  title        = {Pith review of: Motion Control in Multi-Rotor Aerial Robots Using Deep Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4RI4U5J5}},
  note         = {Machine review of arXiv:2502.05996}
}
read the original abstract

This paper investigates the application of Deep Reinforcement (DRL) Learning to address motion control challenges in drones for additive manufacturing (AM). Drone-based additive manufacturing promises flexible and autonomous material deposition in large-scale or hazardous environments. However, achieving robust real-time control of a multi-rotor aerial robot under varying payloads and potential disturbances remains challenging. Traditional controllers like PID often require frequent parameter re-tuning, limiting their applicability in dynamic scenarios. We propose a DRL framework that learns adaptable control policies for multi-rotor drones performing waypoint navigation in AM tasks. We compare Deep Deterministic Policy Gradient (DDPG) and Twin Delayed Deep Deterministic Policy Gradient (TD3) within a curriculum learning scheme designed to handle increasing complexity. Our experiments show TD3 consistently balances training stability, accuracy, and success, particularly when mass variability is introduced. These findings provide a scalable path toward robust, autonomous drone control in additive manufacturing.

Figures

Figures reproduced from arXiv: 2502.05996 by the authors.

Figure 1
Figure 1. 3D Scatter Plot of Reward Function where µ is the mean of each state variable and σ is the standard deviation of each state variable. The RL agent controls the multirotor using three actions: roll (ϕa), pitch (θa), and thrust (T). Yaw control is omitted in this configuration. Excluding yaw control reduces the complexity of the problem and allows the RL agent to stabilize and control the roll and pitch axes effective… view at source ↗
Figure 2
Figure 2. Simulink Model termination function, designed to enable real-time adaptive learning for stable control. The Simulink environment functions in a continuous loop. At each step, the RL agent processes the latest observations to generate control actions. These actions are scaled, with roll and pitch constrained between − π 2 and π 2 , and thrust limited between 0 and 10. The scaled actions are then processed through the… view at source ↗
Figure 3
Figure 3. Training Comparison of TD3 and DDPG randomly to expose the agent to a more diverse set of condi￾tions. This approach helped the agent to become more robust by training in a variety of environmental configurations, ensuring its generalization ability across different scenarios. A. Performance on Target-Reaching Task The performance of TD3 and DDPG was analyzed using their training progress, as shown in [PITH_FULL_IM… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Test Results of TD3 Agent Trained with Different Methods [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Final Demonstration of Drone Waypoint Navigation [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 31 canonical work pages

  1. [1]

    S. A. H. Mohsan, M. A. Khan, F. Noor, I. Ullah, and M. H. Alsharif, ”Towards the unmanned aerial vehicles (UA Vs): A comprehensive review,” Drones, vol. 6, no. 6, p. 147, 2022

  2. [2]

    Aghaee, L

    K. Aghaee, L. Li, A. Roshan, and P. Namakiaraghi, ”Additive man- ufacturing evolution in construction: from individual terrestrial to collective, aerial, and extraterrestrial applications,” Journal of Building Engineering, p. 110389, 2024

  3. [3]

    Stamatopoulos, A

    M.-N. Stamatopoulos, A. Banerjee, and G. Nikolakopoulos, ”Collab- orative Aerial 3D Printing: Leveraging UA V Flexibility and Mesh Decomposition for Aerial Swarm-Based Construction,” in 2024 Inter- national Conference on Unmanned Aircraft Systems (ICUAS) , 2024: IEEE, pp. 45-52

  4. [4]

    S. A. H. Mohsan, N. Q. H. Othman, Y . Li, M. H. Alsharif, and M. A. Khan, ”Unmanned aerial vehicles (UA Vs): Practical aspects, applications, open challenges, security issues, and future trends,” Intelligent Service Robotics , vol. 16, no. 1, pp. 109-137, 2023

  5. [5]

    Chermprayong, ”Enabling Technologies for Precise Aerial Manu- facturing with Unmanned Aerial Vehicles,” Imperial College London, 2019

    P. Chermprayong, ”Enabling Technologies for Precise Aerial Manu- facturing with Unmanned Aerial Vehicles,” Imperial College London, 2019

  6. [6]

    M. A. Alandihallaj, M. Ramezani, and A. M. Hein, ”MBSE-Enhanced LSTM Framework for Satellite System Reliability and Failure Predic- tion,” in Proceedings of the 12th International Conference on Model- Based Software and Systems Engineering (MODELSWARD), 2024, pp. 349–356. [Online]. Available: https://www.scitepress.org/ Papers/2024/126076/126076.pdf

  7. [7]

    L. G. Kraft and D. P. Campagna, ”A comparison between CMAC neural network control and two traditional adaptive control systems,” IEEE Control Systems Magazine , vol. 10, no. 3, pp. 36-43, 1990

  8. [8]

    Stochastic model predictive control-based countermeasure methodology for satellites against indirect kinetic cyber-attacks,

    M. A. Alandihallaj, N. Assadian, and K. Khorasani, “Stochastic model predictive control-based countermeasure methodology for satellites against indirect kinetic cyber-attacks,” International Journal of Con- trol, vol. 96, no. 7, pp. 1895–1908, 2023

Show all 32 references
  1. [9]

    Patchett, ”On the derivation and analysis of decision architectures for unmanned aircraft systems,” 2013

    C. Patchett, ”On the derivation and analysis of decision architectures for unmanned aircraft systems,” 2013

  2. [10]

    Sliding mode control of electro- magnetic tethered satellite formation,

    M. A. A. Hallaj and N. Assadian, “Sliding mode control of electro- magnetic tethered satellite formation,” Advances in Space Research , vol. 58, no. 4, pp. 619–634, 2016

  3. [11]

    Ramezani and M

    M. Ramezani and M. Amiri Atashgah, ”Energy-Aware Hierarchical Reinforcement Learning Based on the Predictive Energy Consump- tion Algorithm for Search and Rescue Aerial Robots in Unknown Environments,” Drones, vol. 8, no. 7, p. 283, 2024

  4. [12]

    Ramezani, M

    M. Ramezani, M. A. Alandihallaj, and A. M. Hein, ”PPO-Based Dynamic Control of Uncertain Floating Platforms in Zero-G Envi- ronment,” in 2024 IEEE International Conference on Robotics and Automation (ICRA), 2024: IEEE, pp. 11730-11736

  5. [13]

    Ramezani, M

    M. Ramezani, M. A. Alandihallaj, and A. M. Hein, ”Fuel-Efficient and Fault-Tolerant CubeSat Orbit Correction via Machine Learning-Based Adaptive Control,” Aerospace, vol. 11, no. 1, p. 1, 2024. [Online]. Available: https://www.mdpi.com/2226-4310/11/1/1

  6. [14]

    Ramezani, H

    M. Ramezani, H. Habibi, J. L. Sanchez-Lopez, and H. V oos, ”UA V path planning employing MPC-reinforcement learning method con- sidering collision avoidance,” in 2023 International Conference on Unmanned Aircraft Systems (ICUAS) , 2023: IEEE, pp. 507-514

  7. [15]

    Ramezani, M

    M. Ramezani, M. Amiri Atashgah, and A. Rezaee, ”A Fault- Tolerant Multi-Agent Reinforcement Learning Framework for Un- manned Aerial Vehicles–Unmanned Ground Vehicle Coverage Path Planning,” Drones, vol. 8, no. 10, p. 537, 2024

  8. [16]

    H. Tan, ”Reinforcement learning with deep deterministic policy gradi- ent,” in 2021 International Conference on Artificial Intelligence, Big Data and Algorithms (CAIBDA) , 2021: IEEE, pp. 82-85

  9. [17]

    Teixeira, G

    K. Teixeira, G. Miguel, H. S. Silva, and F. Madeiro, ”A survey on applications of unmanned aerial vehicles using machine learning,” IEEE Access, 2023

  10. [18]

    Song et al., ”From deterministic to stochastic: an interpretable stochastic model-free reinforcement learning framework for portfolio optimization,” Applied Intelligence, vol

    Z. Song et al., ”From deterministic to stochastic: an interpretable stochastic model-free reinforcement learning framework for portfolio optimization,” Applied Intelligence, vol. 53, no. 12, pp. 15188-15203, 2023

  11. [19]

    Human-centric aware UA V trajectory planning in search and res- cue missions employing multi-objective reinforcement learning with AHP and similarity-based experience replay,

    M. Ramezani, M. Atashgah, J. L. Sanchez-Lopez, and H. V oos, “Human-centric aware UA V trajectory planning in search and res- cue missions employing multi-objective reinforcement learning with AHP and similarity-based experience replay,” in 2024 International Conference on Unm...

  12. [20]

    Towards autonomous multi-UA V wireless network: A survey of reinforcement learning-based approaches,

    Y . Bai, H. Zhao, X. Zhang, Z. Chang, R. J¨antti, and K. Yang, “Towards autonomous multi-UA V wireless network: A survey of reinforcement learning-based approaches,” IEEE Communications Surveys & Tutori- als, 2023

  13. [21]

    Machine learning-aided operations and communications of unmanned aerial vehicles: A contemporary survey,

    H. Kurunathan, H. Huang, K. Li, W. Ni, and E. Hossain, “Machine learning-aided operations and communications of unmanned aerial vehicles: A contemporary survey,” IEEE Communications Surveys & Tutorials, 2023

  14. [22]

    A survey on curriculum learning,

    X. Wang, Y . Chen, and W. Zhu, “A survey on curriculum learning,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 9, pp. 4555–4576, 2021

  15. [23]

    Safe and adaptive autonomous navigation under uncertainty based on sequen- tial waypoints and reachability analysis,

    N. M. B. Lakhal, L. Adouane, O. Nasri, and J. B. H. Slama, “Safe and adaptive autonomous navigation under uncertainty based on sequen- tial waypoints and reachability analysis,” Robotics and Autonomous Systems, vol. 152, p. 104065, 2022

  16. [24]

    Markov decision processes,

    M. L. Puterman, “Markov decision processes,” in Handbooks in Operations Research and Management Science , vol. 2, pp. 331–434, 1990

  17. [25]

    ”Trajectory Generation and Control for Precise Aggressive Maneuvers with Quadrotors.” The International Journal of Robotics Research , 2012, pp

    Mellinger, Daniel, and Nathan Michael. ”Trajectory Generation and Control for Precise Aggressive Maneuvers with Quadrotors.” The International Journal of Robotics Research , 2012, pp. 664–674

  18. [26]

    Aerial additive manu- facturing with multiple autonomous robots,

    K. Zhang, P. Chermprayong, F. Xiao, et al., “Aerial additive manu- facturing with multiple autonomous robots,” *Nature*, vol. 609, pp. 709–717, 2022

  19. [27]

    Challeng- ing common assumptions in convex reinforcement learning,

    M. Mutti, R. De Santi, P. De Bartolomeis, and M. Restelli, “Challeng- ing common assumptions in convex reinforcement learning,”Advances in Neural Information Processing Systems , vol. 35, pp. 4489–4502, 2022

  20. [28]

    Pulse-width modulation,

    J. Sun, “Pulse-width modulation,” in Dynamics and Control of Switched Electronic Systems: Advanced Perspectives for Modeling, Simulation and Control of Power Converters , Springer, 2012, pp. 25– 61

  21. [29]

    A novel DDPG method with prioritized experience replay,

    Y . Hou, L. Liu, Q. Wei, X. Xu, and C. Chen, “A novel DDPG method with prioritized experience replay,” in2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC), 2017: IEEE, pp. 316–321

  22. [30]

    Real-time au- tonomous residential demand response management based on twin delayed deep deterministic policy gradient learning,

    Y . Ye, D. Qiu, H. Wang, Y . Tang, and G. Strbac, “Real-time au- tonomous residential demand response management based on twin delayed deep deterministic policy gradient learning,” Energies, vol. 14, no. 3, p. 531, 2021

  23. [31]

    Lin, ”Self-improving reactive agents based on reinforcement learning, planning, and teaching,” Machine Learning , vol

    L.-J. Lin, ”Self-improving reactive agents based on reinforcement learning, planning, and teaching,” Machine Learning , vol. 8, no. 3- 4, pp. 293-321, 1992

  24. [32]

    Leonid Datta, A Survey on Activation Functions and their relation with Xavier and He Normal Initialization , arXiv, 2020, https://arxiv.org/abs/2004.06632, doi: 10.48550/ARXIV .2004.06632

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.