Pith. sign in

REVIEW 4 major objections 5 minor 61 references

A Self-Supervised Robotic System for Autonomous Contact-Based Spatial Mapping of Semiconductor Properties

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A self-supervised robotic probe predicts pixel-precise contact points without labeled training data and maps photoconductivity across perovskite films at over 125 measurements per hour.

desk verdict A real 24-hour autonomy demo with a genuinely self-supervised pose loss, but the headline 20% gain is unproven because the baselines are trained on the same loss the SDCNN optimizes. read the letter →

arxiv 2411.09892 v2 pith:MG4IGVYP submitted 2024-11-15 cs.RO cond-mat.mtrl-scics.LG

Alexander E. Siemenn , Basita Das , Kangyu Ji , Fang Sheng , Tonio Buonassisi This is my paper · ORCID
classification cs.ROcond-mat.mtrl-scics.LG
keywords autonomousroboticsposepredictionself-supervisedlearningspatialdifferentiabilitypathplanninghigh-throughputcharacterizationcontact-basedperovskitephotoconductivity
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 claims that a contact-based robotic characterization system can be made fully autonomous, without labeled training data, by embedding a domain expert's measurement rules directly into a neural network loss function. The authors design a spatially differentiable CNN (SDCNN) that uses segmented images of drop-cast films as shape priors, optimizing predicted probe poses for coverage, non-overlap, and angular spread. They pair it with a noisy Dijkstra path planner that produces short, low-variance travel routes between poses. Together, they report a 20.0% improvement in valid pose generation over robust loss baselines, and demonstrate a 24-hour unattended run that records 3,025 photoconductivity measurements at over 125 per hour on perovskite films. The resulting spatial maps resolve composition trends and identify inhomogeneities, which matters because high-throughput synthesis in self-driving laboratories needs equally fast, reliable characterization.

What carries the argument

The load-bearing mechanism is the spatially differentiable loss function. It takes the FastSAM image segment $I$ and the predicted poses $\{\mathrm{Pose}_i\}_{i=1}^k$ as shape priors, smooths them with a 2D Gaussian of standard deviation $\sigma$, and composes pose pixels onto segment pixels via an exponential-and-sigmoid operation, making all pixel-space computations back-propagatable. The loss maximizes the covered pixels inside the film and the variance of the yaw angles, subject to pairwise non-overlap; the weights $\omega_1$ and $\omega_2$ are set equal. This transforms unsupervised segmentation into self-supervised training because the image itself provides the error signal. The second mechanism is the noisy Dijkstra planner: it solves an Open Loop Traveling Salesman Problem by building the graph with Dijkstra-style shortest edges while adding i.i.d. uniform noise $\varepsilon_{ij} \sim U(-\alpha d_{ij}, \alpha d_{ij})$ to each edge, then minimizing over generations, which removes the loops and overlaps that plague deterministic planners on clustered pose graphs.

What would settle it

Run the 24-hour campaign on a set of deliberately concave or sharp-cornered films that lie outside the convex, rounded training distribution; if the valid-pose rate falls to chance or poses drift outside the film boundaries, the claimed generality of spatial differentiability fails. Alternatively, insert a known calibration offset of, say, 0.5 mm after hour 12 and check whether the spatial photoconductivity map shifts by that amount; if it does not, the pixel-to-physical correspondence is not the mechanism producing the reported maps.

Watch

Extended reading notes

Core claim

The central claim is that self-supervision via spatial differentiability converts an unsupervised pose-selection problem into a trainable one, producing pixel-precise contact poses without human labels. The SDCNN's loss operates in image space: it smooths the segmented film and predicted poses with Gaussian filters, composes them with a sigmoid, and maximizes the weighted sum of in-film coverage and angular variance subject to a non-overlap constraint. Compared with seven CNN baselines using MSE, MAE, Huber, Poisson, Wing, Reverse Huber, and Barron losses, the SDCNN achieves a 20.0% higher median rate of valid pose generation and 1.5% better positional accuracy with equivalent rotational accuracy and inference time. The noisy Dijkstra planner, which adds uniform noise to edge lengths before minimizing total distance, shortens median path length by 5.0% and tightens variance by 529.1% relative to the best literature planners. In the 24-hour demonstration, 3,025 unique poses were predicted and measured across a MAPb(Br1-xIx)3 composition gradient, yielding photoconductivity maps that show the expected bandgap trend and reveal non-uniform regions attributed to defects or early degradation.

Load-bearing premise

That the manual 15-point image-to-robot calibration, with its correction mesh, stays accurate enough over a 24-hour run that the pixel-predicted poses land on the intended film locations; if it drifts, the photoconductivity values get assigned to the wrong spatial positions, invalidating the spatial maps.

Editorial extensions

If this is right

  • The SDCNN can be applied to other contact-based end effectors: the paper demonstrates both photoconductivity and surface profilometry with the same driving model.
  • The demonstrated throughput of over 125 measurements per hour for 24 hours makes contact-based spatial characterization practical for screening the large composition spaces produced by automated synthesis.
  • Spatial maps produced automatically can flag manufacturing defects such as pinholes and early degradation without human inspection.
  • The inference-time overhead of the proposed loss is only 2.4 ns relative to the slowest baseline, so the accuracy gain comes at no practical compute cost.

Reading between the lines

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

  • If the manual 15-point calibration step were automated with vision-based or learned calibration, the system's 24-hour stability would likely improve further; the authors themselves identify manual calibration as a source of experimental variance.
  • The method's reliance on convex shape priors suggests it will transfer to other drop-cast or printed convex films, but not to highly concave or anisotropic geometries; a conditional pose head (e.g., mixture of experts or transformer) would be needed for arbitrary shapes.
  • The noisy Dijkstra planner's stochastic edge perturbation is a general scheduling trick that could be applied to other clustered routing tasks in laboratory automation, such as pipetting or imaging routes.
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

4 major / 5 minor

Summary. The paper proposes a self-supervised convolutional neural network (SDCNN) with a spatially differentiable loss function for predicting robot contact poses on drop-cast semiconductor films, and a noisy Dijkstra planner for open-loop TSP routing. The system is demonstrated in a 24-hour autonomous photoconductivity mapping campaign collecting 3,025 measurements. The paper reports a 20.0% improvement in valid pose generation and a 1.5% improvement in positional accuracy over robust loss baselines, and a 5.0% shorter median path with ~6x tighter variance compared with other planners.

Significance. If the quantitative claims are correct, the work offers a practical, label-free route to high-throughput contact-based characterization and would be a useful contribution to self-driving laboratories. The 24-hour, 3,025-measurement demonstration is substantial, and the public release of code and data is a strength. However, the central comparison between the SDCNN and baseline losses is self-referential, and the positional accuracy metric lacks a defined ground truth, so the headline improvements require additional independent validation before the claims can be accepted.

major comments (4)
  1. [Methods, Eq. (7) and Fig. 4c] The 20.0% valid-pose improvement is evaluated in a circular manner: the baseline CNNs are trained on labels produced by a random search (N=100) that minimizes the same spatially differentiable loss (Eq. 2) that the SDCNN optimizes directly, and the test metric 'valid pose generation' is the geometric condition embedded in that loss. Because the baselines are regressors trained to imitate this weak stochastic labeler, the reported advantage may be an artifact of the label-generation process rather than a genuine gain of the SDCNN. Please provide an independent ground truth for pose quality (e.g., expert-annotated poses or robot-measured contact success) and either strengthen the labeler (e.g., more random samples or an optimization-based labeler) or use a separate evaluation metric not derived from Eq. (2).
  2. [Fig. 2c and Methods 'Neural network architecture and training'] The reported 'positional accuracy' is never defined; no reference ground truth is given for the predicted pose coordinates. Without knowing whether the reference is the segment boundary, the minimum-loss pose, or a physical measurement, the 1.5% improvement claim cannot be assessed. Specify the ground truth and the error metric used in Fig. 2c.
  3. [Results 'Reliable graph-based robot path planning' and Methods 'Path planning experiments'] The planning baselines are weak or mismatched: A* is a point-to-point shortest-path algorithm, not an OTSP solver; the Dijkstra implementation described is essentially a greedy nearest-neighbor heuristic; and the GA is a single hand-tuned configuration. For graphs of only 105 nodes, standard TSP heuristics (e.g., nearest-neighbor with 2-opt, Christofides with local search) or an exact solver (e.g., Concorde) would be more meaningful comparators. Without such baselines, the 5.0% median path-length improvement and the variance reduction claim are not convincingly established.
  4. [Discussion, calibration limitation (also Fig. S-3)] The manual image-to-robot calibration, which the authors acknowledge is user-skill-dependent and a source of experimental variance, underpins the assignment of the 3,025 photoconductivity measurements to spatial locations in Fig. 6. No evidence is provided that the calibration remains stable over the 24-hour run or that calibration error does not shift the measured values relative to the film positions. Add a repeatability or drift assessment for the calibration (e.g., re-calibration before and after the run, or repeated measurement of a fiducial) and quantify how calibration error propagates into the spatial maps.
minor comments (5)
  1. [Abstract and Results] The abstract says 'refining the pixel precision... by 20.0%', but the Results report the 20.0% improvement for valid-pose generation and a separate 1.5% for positional accuracy. Please align the abstract with the actual metrics.
  2. [Fig. 5g] The phrase '529.1% improvement' in variance is ambiguous; if the variance is reduced from a baseline to 34.3 mm^2, state the baseline variance and the ratio explicitly.
  3. [Eq. (5)-(6)] The Gaussian standard deviation sigma is not reported; this parameter directly controls the definition of a valid pose, so please state the value used in training and evaluation.
  4. [Fig. 4c (right) vs Fig. 5g] Inference time is compared on an NVIDIA Tesla V100 GPU, while the planning experiments use an RTX 4090; ensure the hardware differences are clearly separated for the two comparisons.
  5. [Supplementary Fig. S-5] The model's sensitivity to non-convex shapes is discussed, but the main text does not note the convexity assumption in the definition of validity; add a sentence in the Methods or Results.

Circularity Check

1 steps flagged · score 4.0 of 10

Headline 20% valid-pose improvement rests on a self-referential benchmark: the baselines are trained on stochastic argmin labels from the same Eq. 2 loss that defines both the SDCNN objective and the validity metric.

  1. fitted input called prediction [Methods: 'Neural network architecture and training' (Eq. 7); Results: 'Spatial differentiability for optimal robot pose prediction' (Eq. 2, validity definition); Fig. 4c]
    "Each label is generated by inputting N = 100 randomly generated poses into Eq. 2 for every image. The pose with the lowest loss, Posei∗, becomes the image label for training: i∗ = arg mini∈{1,...,N} loss (I, Posei). ... A predicted pose is considered a valid contact if all its pixels fall within the measurable area (non-zero valued region shown in Fig. 3b) of the differentiable segment I′"

    The SDCNN's training signal (Eq. 2) is a soft, differentiable version of the same geometric condition used to score 'valid pose generation' at test time: all pose pixels must lie inside the segmented film. The seven baseline CNNs are not trained on independent ground-truth expert poses; their labels are obtained by taking the argmin of the same Eq. 2 over N=100 random poses. The reported 20.0% improvement therefore measures how much better a direct optimizer of a metric is than a regressor trained to imitate a weak stochastic approximation of that identical metric. The comparison is partly a property of the label-generation process (N=100 random search) rather than an independent validation of pose quality.

full rationale

The paper is not globally circular: the physical demonstration—3,025 photoconductivity measurements in 24 hours at over 125 measurements per hour, the composition-dependent Gph trend consistent with the MAPbBr3-to-MAPbI3 bandgap trend, and the spatial maps—is independent of the inter-model pose comparison and provides real-world grounding. The noisy Dijkstra planner is benchmarked against external algorithms (A*, Christofides, Dijkstra, GA) and does not reduce to its own inputs. The main circularity is confined to the headline valid-pose claim: the evaluation metric is a thresholded version of the SDCNN's own loss, and the baselines are trained on labels generated by random search optimizing that same loss. This makes the claimed 20.0% improvement over baselines partly by construction, since the baselines are handicapped by a weak labeler rather than by a fundamentally different task. However, the SDCNN does genuinely optimize and generalize the objective, and the physical 24-hour run is independent evidence, so the paper is only partially circular rather than fully self-referential.

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

The performance claims rest on hand-set hyperparameters (loss weights, Gaussian sigma, planner noise and generation count) and on domain assumptions about what makes a pose valid or optimal. The pose-quality evaluation is defined internally through the same loss function used for training, so the ledger flags the objective and the calibration as the main external dependencies.

free parameters (4)
  • omega_1 and omega_2 (loss weights) = set equal (omega_1 = omega_2)
    Balance coverage and angular-variance objectives in Eq. 2; no sensitivity analysis reported.
  • sigma (Gaussian filter std dev) = not specified numerically in main text
    Controls differentiability and soft validity margin in Eqs. 5-6; affects which poses count as valid; tuned by hand.
  • alpha (noise level in noisy Dijkstra) = 0.02
    Hand-tuned noise parameter for edge distances (Eq. 3); no sensitivity study reported.
  • generations for noisy Dijkstra = 1000
    Hand-tuned iteration count for path planning; paired with alpha; no sensitivity analysis reported.
assumptions (4)
  • domain assumption Maximizing coverage, angular variance, and non-overlap of probe poses defines an optimal measurement plan.
    Introduced in Eq. 2 as domain information; the SDCNN is trained to satisfy it, but no external evidence links this objective to measurement quality.
  • domain assumption FastSAM segmentation accurately identifies film boundaries.
    Segments I are used as shape priors; errors would move valid poses outside the film.
  • ad hoc to paper Gaussian smoothing of segments and poses gives a good differentiable approximation of the hard validity constraint.
    Eqs. 5-6; sigma is tunable and not reported, so the soft-to-hard gap is unquantified.
  • domain assumption A 15-point manual calibration mesh is sufficient to map image coordinates to robot coordinates.
    Fig. S-3; the authors note calibration is manual and a source of variance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Self-Supervised Robotic System for Autonomous Contact-Based Spatial Mapping of Semiconductor Properties." pith.science (2026). https://pith.science/paper/MG4IGVYP

@misc{pith2026241109892,
  author       = {Pith},
  title        = {Pith review of: A Self-Supervised Robotic System for Autonomous Contact-Based Spatial Mapping of Semiconductor Properties},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MG4IGVYP}},
  note         = {Machine review of arXiv:2411.09892}
}
read the original abstract

Integrating robotically driven contact-based material characterization techniques into self-driving laboratories can enhance measurement quality, reliability, and throughput. While deep learning models support robust autonomy, current methods lack reliable pixel-precision positioning and require extensive labeled data. To overcome these challenges, we propose an approach for building self-supervised autonomy into contact-based robotic systems that teach the robot to follow domain expert measurement principles at high-throughputs. Firstly, we design a vision-based, self-supervised convolutional neural network (CNN) architecture that uses differentiable image priors to optimize domain-specific objectives, refining the pixel precision of predicted robot contact poses by 20.0% relative to existing approaches. Secondly, we design a reliable graph-based planner for generating distance-minimizing paths to accelerate the robot measurement throughput and decrease planning variance by 6x. We demonstrate the performance of this approach by autonomously driving a 4-degree-of-freedom robotic probe for 24 hours to characterize semiconductor photoconductivity at 3,025 uniquely predicted poses across a gradient of drop-casted perovskite film compositions, achieving throughputs over 125 measurements per hour. Spatially mapping photoconductivity onto each drop-casted film reveals compositional trends and regions of inhomogeneity, valuable for identifying manufacturing process defects. With this self-supervised CNN-driven robotic system, we enable high-precision and reliable automation of contact-based characterization techniques at high throughputs, thereby allowing the measurement of previously inaccessible yet important semiconductor properties for self-driving laboratories.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 59 canonical work pages

  1. [1]

    & Masetti, E

    Piegari, A. & Masetti, E. Thin film thickness measurement: a comparison of various techniques. Thin solid films 124, 249–257 (1985)

    work page 1985

  2. [2]

    Brown, C. A. & Savary, G. Describing ground surface texture using contact profilometry and fractal analysis. Wear 141, 211–226 (1991)

    work page 1991

  3. [3]

    J., Bicanic, N

    Zhu, W., Hughes, J. J., Bicanic, N. & Pearce, C. J. Nanoindentation map- ping of mechanical properties of cement paste and natural rocks. Materials characterization 58, 1189–1198 (2007)

    work page 2007

  4. [4]

    Minor, A. M. et al. A new view of the onset of plasticity during the nanoindentation of aluminium. Nature materials 5, 697–702 (2006)

    work page 2006

  5. [5]

    & Morita, S

    Custance, O., Perez, R. & Morita, S. Atomic force microscopy as a tool for atom manipulation. Nature nanotechnology 4, 803–810 (2009)

    work page 2009

  6. [6]

    Su, J. et al. Intelligent synthesis of magnetic nanographenes via chemist-intuited atomic robotic probe. Nature Synthesis 3, 466–476 (2024)

    work page 2024

  7. [7]

    Ebbesen, T. et al. Electrical conductivity of individual carbon nanotubes. Nature 382, 54–56 (1996)

    work page 1996

  8. [8]

    Wang, Y. et al. Probing photoelectrical transport in lead halide perovskites with van der waals contacts. Nature Nanotechnology 15, 768–775 (2020)

    work page 2020

Show all 61 references
  1. [9]

    Bash, D. et al. Accelerated automated screening of viscous graphene suspensions with various surfactants for optimal electrical conductivity. Digital Discovery 1, 139–146 (2022)

    2022

  2. [10]

    Chandra, H. et al. Open-source automated mapping four-point probe. Materials 10, 110 (2017)

    2017

  3. [11]

    Shimanovich, K. et al. Four-point probe electrical resistivity scanning system for large area conductivity and activation energy mapping. Review of scientific instruments 85 (2014)

    2014

  4. [12]

    & Bonaccurso, E

    Sun, L., Wang, J. & Bonaccurso, E. Conductivity of individual particles measured by a microscopic four-point-probe method. Scientific reports 3, 1991 (2013)

    2013

  5. [13]

    & Dastres, R

    Soori, M., Arezoo, B. & Dastres, R. Artificial intelligence, machine learning and deep learning in advanced robotics, a review. Cognitive Robotics 3, 54–70 (2023)

    2023

  6. [14]

    I., Balter, M

    Chen, A. I., Balter, M. L., Maguire, T. J. & Yarmush, M. L. Deep learning robotic guidance for autonomous vascular access. Nature Machine Intelligence 2, 104–115 (2020). 19

    2020

  7. [15]

    Hippalgaonkar, K. et al. Knowledge-integrated machine learning for materials: lessons from gameplaying and robotics. Nature Reviews Materials 8, 241–260 (2023)

    2023

  8. [16]

    & Benhabib, B

    Nejat, G. & Benhabib, B. High-precision task-space sensing and guidance for autonomous robot localization. 2003 IEEE International Conference on Robotics and Automation 1, 1527–1532 (2003)

    2003

  9. [17]

    & Topakci, M

    ¨Unal, ˙I. & Topakci, M. International Journal of Advanced Robotic Systems 12, 194 (2015)

    2015

  10. [18]

    & Qiao, H

    Li, R. & Qiao, H. A survey of methods and strategies for high-precision robotic grasping and assembly tasks—some new trends. IEEE/ASME Transactions on Mechatronics 24, 2718–2732 (2019)

    2019

  11. [19]

    J., Gauthier, M

    Leveziel, M., Haouas, W., Laurent, G. J., Gauthier, M. & Dahmouche, R. Migri- bot: A miniature parallel robot with integrated gripping for high-throughput micromanipulation. Science Robotics 7, eabn4292 (2022)

    2022

  12. [20]

    Kuwata, Y. et al. Real-time motion planning with applications to autonomous urban driving. IEEE Transactions on control systems technology 17, 1105–1118 (2009)

    2009

  13. [21]

    The limits and potentials of deep learning for robotics

    S¨ underhauf, N.et al. The limits and potentials of deep learning for robotics. The International journal of robotics research 37, 405–420 (2018)

    2018

  14. [22]

    C., Greenewald, K., Lee, K

    Thompson, N. C., Greenewald, K., Lee, K. & Manso, G. F. The computational limits of deep learning. arXiv preprint arXiv:2007.05558 10 (2020)

    2020 arXiv

  15. [23]

    T., Bremer, B

    Rapp, J. T., Bremer, B. J. & Romero, P. A. Self-driving laboratories to autonomously navigate the protein fitness landscape. Nature chemical engineering 1, 97–107 (2024)

    2024

  16. [24]

    Wang, P. et al. Data-driven process characterization and adaptive control in robotic arc welding. CIRP Annals 71, 45–48 (2022)

    2022

  17. [25]

    Siemenn, A. E. et al. Using scalable computer vision to automate high-throughput semiconductor characterization. Nature Communications 15, 4654 (2024)

    2024

  18. [26]

    Omidvar, M. et al. Accelerated discovery of perovskite solid solutions through automated materials synthesis and characterization. Nature Communications 15, 6554 (2024)

    2024

  19. [27]

    Zhao, H. et al. A robotic platform for the synthesis of colloidal nanocrystals. Nature Synthesis 2, 505–514 (2023)

    2023

  20. [28]

    Azizi, S. et al. Autonomous hyperspectral characterisation station: Robot aided measuring of polymer degradation. IEEE Transactions on Automation Science 20 and Engineering (2024)

    2024

  21. [29]

    Mahjour, B. et al. Rapid planning and analysis of high-throughput experiment arrays for reaction discovery. Nature Communications 14, 3924 (2023)

    2023

  22. [30]

    A., Muntasa, A

    Siradjuddin, I. A., Muntasa, A. et al. Faster region-based convolutional neural network for mask face detection. 2021 5th international conference on informatics and computational sciences (ICICoS) 282–286 (2021)

    2021

  23. [31]

    & Wu, X.-J

    Feng, Z.-H., Kittler, J., Awais, M., Huber, P. & Wu, X.-J. Wing loss for robust facial landmark localisation with convolutional neural networks. Proceedings of the IEEE conference on computer vision and pattern recognition 2235–2245 (2018)

    2018

  24. [32]

    & Lambert-Lacroix, S

    Zwald, L. & Lambert-Lacroix, S. The berhu penalty and the grouped effect (2012)

    2012

  25. [33]

    Barron, J. T. A general and adaptive robust loss function. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition 4331–4339 (2019)

    2019

  26. [34]

    Screening robotics and automation

    McGee, J. Screening robotics and automation. Journal of Biomolecular Screening 19, 1131–1132 (2014)

    2014

  27. [35]

    Zhao, X. et al. Fast segment anything (2023)

    2023

  28. [36]

    & Rinaldi, G

    J¨ unger, M., Reinelt, G. & Rinaldi, G. The traveling salesman problem.Handbooks in operations research and management science 7, 225–330 (1995)

    1995

  29. [37]

    Chieng, H. H. & Wahid, N. A performance comparison of genetic algorithm’s mutation operators in n-cities open loop travelling salesman problem. Recent Advances on Soft Computing and Data Mining: Proceedings of The First Interna- tional Conference on Soft Computing and Data Min...

    2014

  30. [38]

    Genetic algorithms for the traveling salesman problem

    Potvin, J.-Y. Genetic algorithms for the traveling salesman problem. Annals of Operations Research 63, 337–370 (1996)

    1996

  31. [39]

    & Raphael, B

    Hart, P., Nilsson, N. & Raphael, B. A formal basis for the heuristic determination of minimum cost paths. IEEE Transactions on Systems Science and Cybernetics 4, 100–107 (1968)

    1968

  32. [40]

    Worst-case analysis of a new heuristic for the traveling salesman problem

    Christophides, N. Worst-case analysis of a new heuristic for the traveling salesman problem. Proc. Symposium on New Directions and Recent Results in Algorithms and Complexity (1976)

    1976

  33. [41]

    Dijkstra, E. W. A note on two problems in connexion with graphs. Numerische mathematik 1, 269–271 (1959). 21

    1959

  34. [42]

    MacLeod, B. P. et al. Self-driving laboratory for accelerated discovery of thin-film materials. Science Advances 6, eaaz8867 (2020)

    2020

  35. [43]

    Langner, S. et al. Beyond ternary opv: high-throughput experimentation and self- driving laboratories optimize multicomponent systems. Advanced Materials 32, 1907801 (2020)

    2020

  36. [44]

    Bag, M. et al. Rapid combinatorial screening of inkjet-printed alkyl-ammonium cations in perovskite solar cells. Materials Letters 164, 472–475 (2016)

    2016

  37. [45]

    & Zhang, L

    Luo, S., Li, T., Wang, X., Faizan, M. & Zhang, L. High-throughput computa- tional materials screening and discovery of optoelectronic semiconductors. Wiley Interdisciplinary Reviews: Computational Molecular Science 11, e1489 (2021)

    2021

  38. [46]

    H., Singh, S

    Son, K. H., Singh, S. P. & Sohn, K.-S. Discovery of novel phosphors for use in light emitting diodes using heuristics optimization-assisted combinatorial chemistry. Journal of Materials Chemistry 22, 8505–8511 (2012)

    2012

  39. [47]

    Jang, D. M. et al. Reversible halide exchange reaction of organometal trihalide perovskite colloidal nanocrystals for full-range band gap tuning. Nano letters 15, 5191–5199 (2015)

    2015

  40. [48]

    E., Meyer, L., Buonassisi, T

    Wieghold, S., Morishige, A. E., Meyer, L., Buonassisi, T. & Sachs, E. M. Crack detection in crystalline silicon solar cells using dark-field imaging. Energy Procedia 124, 526–531 (2017)

    2017

  41. [49]

    & Demant, M

    Kunze, P., Rein, S., Hemsendorf, M., Ramspeck, K. & Demant, M. Learning an empirical digital twin from measurement images for a comprehensive quality inspection of solar cells. Solar RRL 6, 2100483 (2022)

    2022

  42. [50]

    scikit-opt: Python library for optimization algorithms

    Guofei9987. scikit-opt: Python library for optimization algorithms. https:// github.com/guofei9987/scikit-opt/ (2023). Accessed: 2024-04-27

    2023

  43. [51]

    & Aburaia, M

    Hofbauer, C., Aburaia, A., Stuja, K. & Aburaia, M. Automatic print bed leveling for industrial robot systems. Annals of DAAAM & Proceedings 34 (2023)

    2023

  44. [52]

    Enebuse, I. et al. A comparative review of hand-eye calibration techniques for vision guided robots. IEEE Access 9, 113143–113155 (2021)

    2021

  45. [53]

    & Luo, X

    Li, Z., Li, S. & Luo, X. Data-driven industrial robot arm calibration: a machine learning perspective. 2021 IEEE International Conference on Networking, Sensing and Control (ICNSC) 1, 1–6 (2021)

    2021

  46. [54]

    Kong, L.-B. & Yu, Y. Precision measurement and compensation of kinematic errors for industrial robots using artifact and machine learning. Advances in Manufacturing 10, 397–410 (2022). 22

    2022

  47. [55]

    Zhu, J. et al. Uni-perceiver-moe: Learning sparse generalist models with conditional moes (2022)

    2022

  48. [56]

    Vaswani, A. et al. Attention is all you need. Advances in Neural Information Processing Systems 30 (2017)

    2017

  49. [57]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Devlin, J. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805 (2018)

    2018 arXiv

  50. [58]

    Han, Y. et al. Dynamic neural networks: A survey.IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 7436–7456 (2021)

    2021

  51. [59]

    P., Mohamed, S., Jimenez Rezende, D

    Kingma, D. P., Mohamed, S., Jimenez Rezende, D. & Welling, M. Semi-supervised learning with deep generative models. Advances in neural information processing systems 27 (2014)

    2014

  52. [60]

    & Oudeyer, P.-Y

    Baranes, A. & Oudeyer, P.-Y. Active learning of inverse models with intrinsically motivated goal exploration in robots. Robotics and Autonomous Systems 61, 49–73 (2013)

    2013

  53. [61]

    & Kweon, I

    Woo, S., Park, J., Lee, J.-Y. & Kweon, I. S. Cbam: Convolutional block attention module. Proceedings of the European conference on computer vision (ECCV) 3–19 (2018). Acknowledgements The authors thank Julia Hsu for her contribution to the methodology of this research and Tian...

    2018

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.