arxiv: 2605.14367 · v1 · submitted 2026-05-14 · 📡 eess.SY · cs.HC· cs.SY· math.OC

Recognition: 2 theorem links

· Lean Theorem

Automated Curriculum Design for High-dimensional Human Motor Learning

Ankur Kamboj , Rajiv Ranganathan , Xiaobo Tan , Vaibhav Srivastava

Authors on Pith no claims yet

Pith reviewed 2026-05-15 02:31 UTC · model grok-4.3

classification 📡 eess.SY cs.HCcs.SYmath.OC

keywords curriculum designmotor learningmodel predictive controlskill acquisitionhuman subjectsexoskeletonde-novo tasksskill estimation

0 comments

The pith

A framework using motor learning models and stochastic nonlinear MPC designs curricula that speed skill acquisition by about 23 percent over random schedules.

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

The paper develops an automated method for choosing practice tasks in complex motor skills when the learner's true skill state remains hidden from direct observation. It merges a model of human motor learning, continuous estimates of current skill level, and stochastic nonlinear model predictive control to select each next task. Both computer simulations and a study with 36 participants using a hand exoskeleton tested the approach on novel high-dimensional tasks. The resulting curricula produced faster learning than either random task orders or simple rules based only on recent performance. The measured gains reached roughly 23 percent over random curricula and 17 percent over the performance-based alternative.

Core claim

The paper shows that an automated curriculum design system built from a human motor learning model and real-time skill estimation, embedded inside stochastic nonlinear model predictive control, selects task sequences that accelerate skill acquisition in de-novo high-dimensional motor tasks. Validation occurred through both numerical simulations and a controlled human-subject experiment with 36 participants operating a hand exoskeleton. The method delivered measurable reductions in time to reach target skill levels compared with random and heuristic baselines.

What carries the argument

Stochastic nonlinear model predictive control that optimizes task selection by predicting future skill evolution from the motor learning model and current skill estimates.

Load-bearing premise

The human motor learning model together with real-time skill estimation correctly infers the hidden internal skill states that the control algorithm needs to choose good tasks.

What would settle it

A new controlled human experiment on a comparable high-dimensional motor task in which the proposed curriculum produces learning rates equal to or slower than a random schedule would falsify the claim.

Figures

Figures reproduced from arXiv: 2605.14367 by Ankur Kamboj, Rajiv Ranganathan, Vaibhav Srivastava, Xiaobo Tan.

**Figure 1.** Figure 1: Target Capture Game: (a) A participant playing the target capture game with the hand exoskeleton (SenseGlove DK1) strapped on the right hand. (b) The cursor trajectories during Block 1, and (c) Block 8 of the participant’s target capture gameplay are shown. The red dots are the target points, and the 5 × 5 unit grid is the task space shown to the participants. The grid units are computed to ensure particip… view at source ↗

**Figure 2.** Figure 2: a shows the efficacy of the designed particle filter in estimating the learning state Cˆ and capturing the learning error trend across trials in simulation. The performance of three curricula was evaluated through simulation first, where a fitted HML model played the target capture game across 10 Monte Carlo repetitions per group. These groups were defined by their scheduling logic: random (control group),… view at source ↗

**Figure 3.** Figure 3: Task and Learning Performance: (a) Learning error evolution for participants playing the game under the three curricula shows that on average, the SNMPC group takes 109 and 80 fewer trials than the control and manual groups, respectively, to achieve 80% learning. (b) SoT curves of participants from the three groups, and (c) the mean number of trials (and the scatter plot across participants and target pair… view at source ↗

**Figure 4.** Figure 4: Uncontrolled Manifold Analysis: The fraction of trial-to-trial variability in the hand finger-joints of the participants projected onto the UCM of the task for (a) phase 1, (b) phase 2 as a function of the normalized trial time, and (c) the total fraction of VUCM across the trials for the two phases under the three sequencing strategies shows that the variability decreases for all groups, an evidence of t… view at source ↗

**Figure 5.** Figure 5: Robustness of SNMPC Curriculum to Model Mismatches: FME curves for fitted participant model A playing the game on SNMPC curriculum, with SNMPC using model A (A-A) or using model B (A-B), show that learning performance degrades (a) for model A-B with increasing lookahead horizon (2, 4, 6), indicating the importance of the correct model in SNMPC. (b) In the case of no model mismatch (Model A-A), increasing τ… view at source ↗

**Figure 6.** Figure 6: Gameplay Trajectories: Gameplay trajectories for the worst (top) and the best (bottom) participants in (a) control group, (b) manual group, and (c) SNMPC group show how the participants trained on SNMPC-based curriculum can achieve straighter trajectories towards the end of the training. The best and worst participants are selected based on the ∥SoT∥2 values across the gameplay trials. D Illustration of Pe… view at source ↗

**Figure 7.** Figure 7: shows how SoT is calculated as the maximum deviation of the cursor trajectory from the straight line joining the start and end points, normalized to the distance between the start and the end points [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗

**Figure 8.** Figure 8: Distribution of final state estimation errors for nonlinear filters over 100 Monte-Carlo simulations of one game trial shows that the (a) EKF has high precision but low accuracy, (b) UKF has low precision and low accuracy, and (c) PF has high precision and high accuracy in estimating the state Wˆ . To demonstrate the efficacy of PF over Gaussian filters for state estimation in our setup, we run a consisten… view at source ↗

read the original abstract

Designing effective practice schedules for high-dimensional motor learning tasks remains a challenge, especially when skill states are unobservable and task performance may not reflect the true learning. We propose an automated curriculum design framework that combines a human motor learning model and personalized real-time skill estimation with Stochastic Nonlinear Model Predictive Control in \emph{de-novo} (novel) motor learning paradigms. We validated our framework both through simulations and human-subject studies (N = 36) using a hand exoskeleton. Our proposed approach accelerates skill acquisition by $\sim23\%$, and ${\sim17\%}$ when compared to a random curriculum and a performance heuristics-based curriculum, respectively. These significant gains in learning efficiency highlight the potential of model-based, individualized curricula for motor rehabilitation and complex skill training.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

The paper integrates a motor learning model, real-time latent skill estimation, and stochastic nonlinear MPC to generate personalized curricula for novel high-dimensional tasks, with human data showing gains over random and heuristic baselines.

read the letter

The core contribution is a closed-loop system that treats skill acquisition as a partially observable process and uses MPC to pick practice sequences that target estimated hidden states rather than raw performance. They test it in simulation and then in a hand-exoskeleton study with 36 participants on a de-novo task, reporting roughly 23 % faster acquisition than random curricula and 17 % faster than a performance-heuristic controller. That combination of model, estimator, and optimizer applied to real human data is the new piece relative to earlier motor-learning or adaptive-training work. The experimental setup is straightforward and the reported effect sizes are large enough to be practically interesting for rehabilitation contexts. The math and control formulation appear standard once the model is accepted, and the human validation moves it beyond pure simulation. The main limitation is that the only outcome measures are the same task-performance curves used to drive the estimator and the controller. There are no retention tests, transfer tasks, or other probes that would show the estimated skill states actually track durable learning rather than transient performance. If the estimator is mainly chasing the observable metric, the curriculum gains could shrink or disappear under different evaluation. The model parameters themselves are fitted rather than derived from first principles, so the framework inherits whatever biases are in that fit. Readers working on model-based adaptive training or exoskeleton-assisted rehab will find the method and the numbers useful. It is not a foundational theoretical advance, but the empirical application is concrete enough that a serious referee should see it. I would send it to review with a request for retention or transfer data in revision.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes an automated curriculum design framework for high-dimensional de-novo motor learning tasks. It integrates a human motor learning model with personalized real-time skill estimation and Stochastic Nonlinear Model Predictive Control (SNMPC) to select individualized practice schedules. Validation consists of simulations plus a human-subject study (N=36) using a hand exoskeleton, with the central claim that the approach accelerates skill acquisition by approximately 23% relative to a random curriculum and 17% relative to a performance-heuristics curriculum.

Significance. If the results hold, the work could meaningfully advance motor rehabilitation and complex skill training by showing that model-based, individualized curricula outperform standard baselines. The combination of a control-theoretic optimizer with human-subject experiments is a clear strength, and the N=36 study provides direct empirical grounding. However, the significance is limited by the absence of independent validation that the estimator recovers unobservable skill states rather than performance artifacts.

major comments (2)

[Human-Subject Validation] Human-Subject Validation section: the N=36 study reports only task performance curves; no retention test, transfer task, or other ground-truth probe is described that would confirm the real-time skill estimator tracks the latent learning state rather than observable performance. Because the SNMPC curriculum selection depends on these estimates, this gap directly undermines the claim that the observed gains reflect accelerated learning rather than metric-specific effects.
[Simulation Validation] Simulation Validation section: simulations assume the human motor learning model is correct by construction and therefore cannot falsify the link between estimated skill states and true learning; they provide no independent support for the human-experiment results that rely on the same unobservable-state premise.

minor comments (1)

[Abstract] Abstract: the reported percentages should be accompanied by the statistical test and p-value used to establish significance, and the exact definition of 'skill acquisition' (e.g., time to reach a performance threshold) should be stated explicitly.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of validating unobservable skill states in our framework. We address each major comment below and have revised the manuscript accordingly where feasible.

read point-by-point responses

Referee: [Human-Subject Validation] Human-Subject Validation section: the N=36 study reports only task performance curves; no retention test, transfer task, or other ground-truth probe is described that would confirm the real-time skill estimator tracks the latent learning state rather than observable performance. Because the SNMPC curriculum selection depends on these estimates, this gap directly undermines the claim that the observed gains reflect accelerated learning rather than metric-specific effects.

Authors: We agree that retention or transfer tests would provide stronger direct evidence that the estimator captures latent learning states rather than performance artifacts. Designing suitable transfer tasks for this high-dimensional de-novo paradigm is nontrivial, which is why the study focused on acquisition-phase performance as the primary metric. The superior outcomes relative to the performance-heuristics baseline (which uses only observable metrics) provide indirect support that the model-based estimates add value. We have added a dedicated paragraph in the revised Discussion section acknowledging this limitation and outlining future work with retention and transfer probes. revision: yes
Referee: [Simulation Validation] Simulation Validation section: simulations assume the human motor learning model is correct by construction and therefore cannot falsify the link between estimated skill states and true learning; they provide no independent support for the human-experiment results that rely on the same unobservable-state premise.

Authors: The simulations are intended only to verify the SNMPC optimizer and curriculum logic under the assumption that the model holds, serving as a controlled proof-of-concept for the control framework. They are not presented as independent validation of the human motor learning model. The N=36 human experiments constitute the primary empirical test. We have clarified this scope and role of the simulations in the revised Simulation Validation section. revision: partial

Circularity Check

0 steps flagged

No significant circularity; central claims rest on external human validation

full rationale

The paper's derivation uses a human motor learning model plus real-time estimator inside stochastic nonlinear MPC to generate curricula. However, the load-bearing performance claims (~23% and ~17% acceleration) are measured directly from observable task performance in an independent N=36 human-subject study on a hand exoskeleton, not recovered from the same fitted parameters or simulations that assume the model. No self-definitional equations, fitted-input predictions, or self-citation chains appear in the abstract or described framework; the human data serve as an external benchmark. The derivation chain is therefore self-contained against those benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The abstract provides no explicit list of free parameters or axioms; the framework necessarily relies on an unspecified human motor learning model whose parameters are presumably fitted or assumed, plus the standard assumptions of stochastic nonlinear MPC.

free parameters (1)

parameters of the human motor learning model
The abstract states that a human motor learning model is combined with the controller; any such model requires fitted parameters for learning rates or state transitions that are not detailed here.

axioms (1)

domain assumption Skill states are unobservable and task performance does not directly reflect true learning progress
Stated in the opening sentence of the abstract as the core motivation for the framework.

pith-pipeline@v0.9.0 · 5439 in / 1331 out tokens · 23976 ms · 2026-05-15T02:31:58.003011+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We propose an automated curriculum design framework that combines a human motor learning model and personalized real-time skill estimation with Stochastic Nonlinear Model Predictive Control
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

ℓj = βW ‖˜Wj‖² + βRE ∥REυ∥² + βSoT ∥SoTυ∥²

What do these tags mean?

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

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · 1 internal anchor

[1]

R. A. Schmidt, T. D. Lee, C. Winstein, G. Wulf, and H. N. Zelaznik, Motor Control and Learning: A Behavioral Emphasis . Human Kinetics, 2018

work page 2018
[2]

Individualized challenge point practice as a method to aid motor sequence learning,

K. P. Wadden, N. J. Hodges, K. L. D. Asis, J. L. Neva, and L. A. Boyd, “Individualized challenge point practice as a method to aid motor sequence learning,” Journal of Motor Behavior , vol. 51, no. 5, pp. 467–485, 2019

work page 2019
[3]

Dynamic diﬀiculty adjustment (DDA) in computer games: A review,

H. Nakanishi and M. Zohaib, “Dynamic diﬀiculty adjustment (DDA) in computer games: A review,” Advances in Human-Computer Interaction , vol. 2018, p. 5681652, 2018

work page 2018
[4]

Learning control in robot-assisted rehabilitation of motor skills: A review,

S.-H. Zhou, J. Fong, V. Crocher, Y. Tan, D. Oetomo, and I. Mareels, “Learning control in robot-assisted rehabilitation of motor skills: A review,” Journal of Control and Decision , vol. 3, no. 1, pp. 19–43, 2016

work page 2016
[5]

Interactive curriculum learning increases and homogenizes motor smoothness,

V. Sungeelee, A. Loriette, O. Sigaud, and B. Caramiaux, “Interactive curriculum learning increases and homogenizes motor smoothness,” Scientific Reports, vol. 14, no. 1, p. 2843, 2024

work page 2024
[6]

Evalu- ating rehabilitation progress using motion features identified by machine learning,

L. Lu, Y. Tan, M. Klaic, M. P. Galea, F. Khan, A. Oliver, I. Mareels, D. Oetomo, and E. Zhao, “Evalu- ating rehabilitation progress using motion features identified by machine learning,” IEEE Transactions on Biomedical Engineering , vol. 68, no. 4, pp. 1417–1428, 2021

work page 2021
[7]

Assessment-driven selection and adaptation of exercise diﬀiculty in robot-assisted therapy: A pilot study with a hand rehabilitation robot,

J.-C. Metzger, O. Lambercy, A. Caliﬀi, D. Dinacci, C. Petrillo, P. Rossi, F. M. Conti, and R. Gassert, “Assessment-driven selection and adaptation of exercise diﬀiculty in robot-assisted therapy: A pilot study with a hand rehabilitation robot,” Journal of NeuroEngineering and Rehabilitation , vol. 11, no. 1, p. 154, 2014

work page 2014
[8]

Capturing skill state in curriculum learning for human skill acquisition,

K. Ghonasgi, R. Mirsky, S. Narvekar, B. Masetty, A. M. Haith, P. Stone, and A. D. Deshpande, “Capturing skill state in curriculum learning for human skill acquisition,” in IEEE/RSJ International Conference on Intelligent Robots and Systems , 2021, pp. 771–776

work page 2021
[9]

Learning–performance distinction and memory processes for motor skills: A focused review and perspective,

S. S. Kantak and C. J. Winstein, “Learning–performance distinction and memory processes for motor skills: A focused review and perspective,” Behavioural Brain Research , vol. 228, no. 1, pp. 219–231, 2012

work page 2012
[10]

Contextual interference effects on the acquisition, retention, and transfer of a motor skill

J. B. Shea and R. L. Morgan, “Contextual interference effects on the acquisition, retention, and transfer of a motor skill. ” Journal of Experimental Psychology: Human Learning and Memory , vol. 5, no. 2, p. 179, 1979

work page 1979
[11]

The myth of contextual interference learning benefit in sports practice: A systematic review and meta- analysis,

A. Ammar, K. Trabelsi, M. A. Boujelbane, O. Boukhris, J. M. Glenn, H. Chtourou, and W. I. Schöllhorn, “The myth of contextual interference learning benefit in sports practice: A systematic review and meta- analysis,” Educational Research Review, vol. 39, p. 100537, 2023. 12

work page 2023
[12]

Contextual interference: A meta-analytic study,

F. Brady, “Contextual interference: A meta-analytic study,” Perceptual and Motor Skills , vol. 99, no. 1, pp. 116–126, 2004

work page 2004
[13]

Performance-based adaptive schedules enhance motor learning,

Y. Choi, F. Qi, J. Gordon, and N. Schweighofer, “Performance-based adaptive schedules enhance motor learning,” Journal of Motor Behavior , vol. 40, no. 4, pp. 273–280, 2008

work page 2008
[14]

A control framework for adaptation of training task and robotic assistance for promoting motor learning with an upper limb rehabilitation robot,

C. Wang, L. Peng, and Z.-G. Hou, “A control framework for adaptation of training task and robotic assistance for promoting motor learning with an upper limb rehabilitation robot,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 52, no. 12, pp. 7737–7747, 2022

work page 2022
[15]

Rowing simulator modulates water density to foster motor learning,

E. Basalp, L. Marchal-Crespo, G. Rauter, R. Riener, and P. Wolf, “Rowing simulator modulates water density to foster motor learning,” Frontiers in Robotics and AI , vol. 6, p. 74, 2019

work page 2019
[16]

V. A. Swanson, C. Johnson, D. K. Zondervan, N. Bayus, P. McCoy, Y. F. J. Ng, B. Jenna Schindele, D. J. Reinkensmeyer, and S. Shaw, “Optimized home rehabilitation technology reduces upper extremity impairment compared to a conventional home exercise program: A randomized, controlled, single-blind trial in subacute stroke,” Neurorehabilitation and Neural Re...

work page 2023
[17]

Human motor learning dynamics in high- dimensional tasks,

A. Kamboj, R. Ranganathan, X. Tan, and V. Srivastava, “Human motor learning dynamics in high- dimensional tasks,” PLOS Computational Biology , vol. 20, no. 10, p. e1012455, 2024

work page 2024
[18]

Expediting human motor learning in high-dimensional de-novo tasks via online curriculum de- sign,

——, “Expediting human motor learning in high-dimensional de-novo tasks via online curriculum de- sign,” in American Control Conference, Denver, CO, 2025, pp. 4641–4646

work page 2025
[19]

Forward models: Supervised learning with a distal teacher,

M. I. Jordan and D. E. Rumelhart, “Forward models: Supervised learning with a distal teacher,” Cognitive Science, vol. 16, no. 3, pp. 307–354, 1992

work page 1992
[20]

An internal model for sensorimotor integration,

D. M. Wolpert, Z. Ghahramani, and M. I. Jordan, “An internal model for sensorimotor integration,” Science, vol. 269, no. 5232, pp. 1880–1882, 1995

work page 1995
[21]

Thrun, W

S. Thrun, W. Burgard, and D. Fox, Probabilistic Robotics. MIT Press, 2005

work page 2005
[22]

The uncontrolled manifold concept: Identifying control variables for a functional task,

J. P. Scholz and G. Schöner, “The uncontrolled manifold concept: Identifying control variables for a functional task,” Experimental Brain Research, vol. 126, no. 3, pp. 289–306, 1999

work page 1999
[23]

Identifying the control structure of multijoint coordination during pistol shooting,

J. P. Scholz, G. Schöner, and M. L. Latash, “Identifying the control structure of multijoint coordination during pistol shooting,” Experimental Brain Research, vol. 135, no. 3, pp. 382–404, 2000

work page 2000
[24]

Win-shift, lose-stay: Contingent switching and contextual interference in motor learning,

D. A. Simon, T. D. Lee, and J. D. Cullen, “Win-shift, lose-stay: Contingent switching and contextual interference in motor learning,” Perceptual and Motor Skills , vol. 107, no. 2, pp. 407–418, 2008

work page 2008
[25]

Challenge point: A framework for conceptualizing the effects of various practice conditions in motor learning,

M. A. Guadagnoli and T. D. Lee, “Challenge point: A framework for conceptualizing the effects of various practice conditions in motor learning,” Journal of Motor Behavior , vol. 36, no. 2, pp. 212–224, 2004

work page 2004
[26]

Toward a new theory of motor synergies,

M. L. Latash, J. P. Scholz, and G. Schöner, “Toward a new theory of motor synergies,” Motor Control, vol. 11, no. 3, pp. 276–308, 2007

work page 2007
[27]

Feed-forward control of a redundant motor system,

S. R. Goodman and M. L. Latash, “Feed-forward control of a redundant motor system,” Biological Cybernetics, vol. 95, no. 3, pp. 271–280, 2006

work page 2006
[28]

Assist-as-needed exoskele- ton for hand joint rehabilitation based on muscle effort detection,

J. C. Castiblanco, I. F. Mondragon, C. Alvarado-Rojas, and J. D. Colorado, “Assist-as-needed exoskele- ton for hand joint rehabilitation based on muscle effort detection,” Sensors, vol. 21, no. 13, p. 4372, 2021

work page 2021
[29]

Subject-specific assist-as-needed controllers for a hand exoskeleton for rehabilitation,

P. Agarwal and A. D. Deshpande, “Subject-specific assist-as-needed controllers for a hand exoskeleton for rehabilitation,” IEEE Robotics and Automation Letters , vol. 3, no. 1, pp. 508–515, 2017. 13

work page 2017
[30]

Robot-assisted adaptive training: Custom force fields for teaching movement patterns,

J. L. Patton and F. A. Mussa-Ivaldi, “Robot-assisted adaptive training: Custom force fields for teaching movement patterns,” IEEE Transactions on Biomedical Engineering , vol. 51, no. 4, pp. 636–646, 2004

work page 2004
[31]

Human-in-the-loop optimization of exoskeleton assistance during walking,

J. Zhang, P. Fiers, K. A. Witte, R. W. Jackson, K. L. Poggensee, C. G. Atkeson, and S. H. Collins, “Human-in-the-loop optimization of exoskeleton assistance during walking,” Science, vol. 356, no. 6344, pp. 1280–1284, 2017

work page 2017
[32]

Learning Models for Shared Control of Human-Machine Systems with Unknown Dynamics

A. Broad, T. Murphey, and B. Argall, “Learning models for shared control of human-machine systems with unknown dynamics,” 2018. [Online]. A vailable: https://arxiv.org/abs/1808.08268

work page internal anchor Pith review Pith/arXiv arXiv 2018
[33]

On human-in-the-loop optimization of human–robot interaction,

P. Slade, C. Atkeson, J. M. Donelan, H. Houdijk, K. A. Ingraham, M. Kim, K. Kong, K. L. Poggensee, R. Riener, M. Steinert et al., “On human-in-the-loop optimization of human–robot interaction,” Nature, vol. 633, no. 8031, pp. 779–788, 2024

work page 2024
[34]

Restricted maximum likelihood (REML) estimation of variance com- ponents in the mixed model,

R. R. Corbeil and S. R. Searle, “Restricted maximum likelihood (REML) estimation of variance com- ponents in the mixed model,” Technometrics, vol. 18, no. 1, pp. 31–38, 1976

work page 1976
[35]

Adjusted p-values for simultaneous inference,

S. P. Wright, “Adjusted p-values for simultaneous inference,” Biometrics, pp. 1005–1013, 1992

work page 1992
[36]

A fast and elitist multiobjective genetic algorithm: NSGA-II,

K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: NSGA-II,” IEEE Transactions on Evolutionary Computation , vol. 6, no. 2, pp. 182–197, 2002. Appendix A Trial-by-trial HML Model The HML model proposed in our previous work [17] is summarized below ˙x = Cu (13a) ˙δq = − aδq + u + ξq (13b) ˙ˆW = − γ ˜W Φδq...

work page 2002