pith. sign in

arxiv: 2604.15854 · v1 · submitted 2026-04-17 · 💻 cs.RO

Limits of Lamarckian Evolution Under Pressure of Morphological Novelty

Jed R Muff , Karine Miras , A.E. Eiben This is my paper

Pith reviewed 2026-05-10 08:41 UTC · model grok-4.3

classification 💻 cs.RO
keywords lamarckianevolutionmorphologicalcontrollerspressuresimilaritydarwiniandiversity
0
0 comments X

The pith

Lamarckian evolution outperforms Darwinian on task performance alone but drops more sharply when morphological novelty is added, because reduced parent-offspring similarity weakens the value of inherited learned controllers.

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

In this setup, robots are built from modules and must learn to move across a surface. Lamarckian evolution lets offspring start with the control programs their parents improved during their own lifetimes. This helps when parent and child bodies are similar. The study first runs evolution only for better movement and finds Lamarckian versions do better than standard Darwinian ones. Then it adds a second goal of making robot shapes more different from each other. Performance falls in both cases, but the drop is larger for Lamarckian evolution. The authors link the extra drop to lower similarity between parents and offspring, which makes the inherited controllers less useful for the new shapes.

Core claim

These results reveal the limits of Lamarckian evolution by exposing a fundamental trade-off between inheritance-based exploitation and diversity-driven exploration.

Load-bearing premise

That the observed performance drop in the Lamarckian condition is primarily caused by reduced parent-offspring morphological similarity rather than other interactions in the multi-objective selection or simulation details.

Figures

Figures reproduced from arXiv: 2604.15854 by A.E. Eiben, Jed R Muff, Karine Miras.

Figure 1
Figure 1. Figure 1: ARIEL framework phenotype space including three [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Representation used in this work. The body employs a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Average locomotion performance (flocomotion) per generation. Higher is better. 0 10 20 30 40 50 Generation 0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Performance Difference (Pure - Combined) Darwinian Lamarckian ↓ [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average difference in locomotion performance. Lower [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Mean parent–child morphological dissimilarity across [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of initial locomotion performance. In [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
read the original abstract

Lamarckian inheritance has been shown to be a powerful accelerator in systems where the joint evolution of robot morphologies and controllers is enhanced with individual learning. Its defining advantage lies in the offspring inheriting controllers learned by their parents. The efficacy of this option, however, relies on morphological similarity between parent and offspring. In this study, we examine how Lamarckian inheritance performs when the search process is driven toward high morphological variance, potentially straining the requirement for parent-offspring similarity. Using a system of modular robots that can evolve and learn to solve a locomotion task, we compare Darwinian and Lamarckian evolution to determine how they respond to shifting from pure task-based selection to a multi-objective pressure that also rewards morphological novelty. Our results confirm that Lamarckian evolution outperforms Darwinian evolution when optimizing task-performance alone. However, introducing selection pressure for morphological diversity causes a substantial performance drop, which is much greater in the Lamarckian system. Further analyses show that promoting diversity reduces parent-offspring similarity, which in turn reduces the benefits of inheriting controllers learned by parents. These results reveal the limits of Lamarckian evolution by exposing a fundamental trade-off between inheritance-based exploitation and diversity-driven exploration.

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.

Circularity Check

0 steps flagged

No significant circularity; purely empirical comparison of evolutionary regimes

full rationale

The paper conducts an experimental study comparing Darwinian and Lamarckian evolution in modular robot locomotion tasks, measuring performance under task-only vs. multi-objective (task + morphological novelty) selection. No derivations, equations, fitted parameters, or predictions are present. Results derive directly from simulation runs and post-hoc similarity analyses without reducing to self-citations, ansatzes, or input-output equivalence. The central claim (trade-off between inheritance exploitation and diversity) follows from observed data rather than any self-referential construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract-only view yields no explicit free parameters or invented entities. The work rests on standard domain assumptions in evolutionary robotics.

axioms (2)
  • domain assumption Modular robot morphologies and controllers can be jointly evolved and learned to solve locomotion tasks
    Core premise of the experimental platform described in the abstract.
  • domain assumption Lamarckian inheritance benefits depend on morphological similarity between parent and offspring
    Invoked to explain why novelty pressure reduces Lamarckian advantage.

pith-pipeline@v0.9.0 · 5513 in / 1200 out tokens · 18367 ms · 2026-05-10T08:41:09.521903+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    Grand chal- lenges for evolutionary robotics,

    A. E. Eiben, S. Kernbach, and E. Haasdijk, “Grand chal- lenges for evolutionary robotics,”Frontiers in Robotics and AI, vol. 2, p. 4, 2015

    work page 2015

  2. [2]

    In: Proceedings of the 2nd International Workshop on Software Health

    M. Jelisavcic, K. Glette, E. Haasdijk, and A. E. Eiben, “Lamarckian evolution of simulated modular robots,” Frontiers in Robotics and AI, vol. 6, p. 9, 2019.DOI: 10.3389/frobt.2019.00009

    work page doi:10.3389/frobt.2019.00009 2019

  3. [3]

    Analysis of lamarckian evolution in morphologically evolving robots,

    M. Jelisavcic, R. Kiesel, K. Glette, E. Haasdijk, and A. E. Eiben, “Analysis of lamarckian evolution in morphologically evolving robots,” inArtificial Life Con- ference Proceedings, MIT Press, 2017, pp. 214–221

    work page 2017

  4. [4]

    Morpho- logical attractors in darwinian and lamarckian evolu- tionary robot systems,

    M. Jelisavcic, K. Miras, and A. E. Eiben, “Morpho- logical attractors in darwinian and lamarckian evolu- tionary robot systems,” in2018 IEEE Symposium Se- ries on Computational Intelligence (SSCI), IEEE, 2018, pp. 859–866

    work page 2018

  5. [5]

    Benefits of lamarckian evolution for morphologically evolving robots,

    M. Jelisavcic, R. Kiesel, K. Glette, E. Haasdijk, and A. E. Eiben, “Benefits of lamarckian evolution for morphologically evolving robots,” inProceedings of the Genetic and Evolutionary Computation Conference Companion, ACM, 2017, pp. 65–66

    work page 2017

  6. [6]

    En- hancing robot evolution through lamarckian principles,

    J. Luo, K. Miras, J. M. Tomczak, and A. E. Eiben, “En- hancing robot evolution through lamarckian principles,” Scientific Reports, vol. 13, no. 1, p. 21 109, 2023

    work page 2023

  7. [7]

    A comparative study of brain reproduction methods for morpholog- ically evolving robots,

    J. Luo, C. Longhi, and A. E. Eiben, “A comparative study of brain reproduction methods for morpholog- ically evolving robots,” inArtificial Life Conference Proceedings, MIT Press, vol. 2023, 2023, p. 44

    work page 2023

  8. [8]

    A comparison of controller architectures and learning mechanisms for arbitrary robot morphologies,

    J. Luo, J. M. Tomczak, K. Miras, and A. E. Eiben, “A comparison of controller architectures and learning mechanisms for arbitrary robot morphologies,” in2023 IEEE Symposium Series on Computational Intelligence (SSCI), IEEE, 2023, pp. 1518–1525

    work page 2023

  9. [9]

    Lamarckian inheritance improves robot evolution in dynamic environments,

    J. Luo, K. Miras, C. Longhi, O. Weissl, and A. E. Eiben, “Lamarckian inheritance improves robot evolution in dynamic environments,”IEEE Transactions on Evolu- tionary Computation, 2025

    work page 2025

  10. [10]

    Investigations into lamarckism, baldwinism and local search in grammatical evolution guided by reinforce- ment,

    J. M. Mingo, R. Aler, D. Maravall, and J. de Lope, “Investigations into lamarckism, baldwinism and local search in grammatical evolution guided by reinforce- ment,”Computing and Informatics, vol. 32, no. 3, pp. 595–627, 2013

    work page 2013

  11. [11]

    Comparison between lamar- ckian and darwinian evolution on a model using neural networks and genetic algorithms,

    T. Sasaki and M. Tokoro, “Comparison between lamar- ckian and darwinian evolution on a model using neural networks and genetic algorithms,”Knowledge and In- formation Systems, vol. 2, no. 2, pp. 201–222, 2000

    work page 2000

  12. [12]

    Developing convolutional neural networks using a novel lamarck- ian co-evolutionary algorithm,

    Z. Sharifi, K. Soltanian, and A. Amiri, “Developing convolutional neural networks using a novel lamarck- ian co-evolutionary algorithm,” in13th International Conference on Computer and Knowledge Engineering (ICCKE), Ferdowsi University of Mashhad, Mashhad, Iran, Nov. 2023

    work page 2023

  13. [13]

    Lamarckian co-design of soft robots via transfer learning,

    N. A. M. Tack, Y . Yamazaki, and F. Iida, “Lamarckian co-design of soft robots via transfer learning,” inPro- ceedings of the Genetic and Evolutionary Computation Conference, ACM, 2024, pp. 129–137

    work page 2024

  14. [14]

    Abandoning objectives: Evolution through the search for novelty alone,

    J. Lehman and K. O. Stanley, “Abandoning objectives: Evolution through the search for novelty alone,”Evolu- tionary Computation, vol. 19, no. 2, pp. 189–223, 2011

    work page 2011

  15. [15]

    Quality diversity: A new frontier for evolutionary computation,

    J. K. Pugh, L. B. Soros, and K. O. Stanley, “Quality diversity: A new frontier for evolutionary computation,” Frontiers in Robotics and AI, vol. 3, p. 40, 2016

    work page 2016

  16. [16]

    Robots that can adapt like animals,

    A. Cully, J. Clune, D. Tarapore, and J.-B. Mouret, “Robots that can adapt like animals,”Nature, vol. 521, no. 7553, pp. 503–507, 2015

    work page 2015

  17. [17]

    Environmental influence on the evolution of morphological complexity in machines,

    J. E. Auerbach and J. C. Bongard, “Environmental influence on the evolution of morphological complexity in machines,”PLOS Computational Biology, vol. 10, no. 1, e1003399, 2014

    work page 2014

  18. [18]

    Un- shackling evolution: Evolving soft robots with multiple materials and a powerful generative encoding,

    N. Cheney, R. MacCurdy, J. Clune, and H. Lipson, “Un- shackling evolution: Evolving soft robots with multiple materials and a powerful generative encoding,”ACM SIGEVOlution, vol. 7, no. 1, pp. 11–23, 2013

    work page 2013

  19. [19]

    Search space analysis of evolvable robot morpholo- gies,

    K. Miras, E. Haasdijk, K. Glette, and A. E. Eiben, “Search space analysis of evolvable robot morpholo- gies,” inApplications of Evolutionary Computation: 21st International Conference, EvoApplications 2018, Parma, Italy, April 4-6, 2018, Proceedings, K. Sim and P. Kaufmann, Eds., ser. Lecture Notes in Computer Sci- ence, vol. 10784, Cham: Springer, 2018, ...

    work page doi:10.1007/978-3-319-77538-8 2018