Pith. sign in

REVIEW 4 major objections 4 minor 49 references

Cyborg Insect Factory: Automatic Assembly System to Build up Insect-computer Hybrid Robot Based on Vision-guided Robotic Arm Manipulation of Custom Bipolar Electrodes

T0 review · 4 major / 4 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 The paper claims that cyborg cockroaches can be assembled automatically in 68 seconds by a vision-guided robotic arm, with steering and deceleration control statistically indistinguishable from manual assembly.

desk verdict A plausible, well-engineered first step toward mass-producing cyborg cockroaches, but the evidence for 'matching' manual assembly is thinner than the abstract suggests and the missing success-rate data is the key gap. read the letter →

arxiv 2411.13164 v1 pith:VMGMMABS submitted 2024-11-20 cs.RO

Qifeng Lin , Nghia Vuong , Kewei Song , Phuoc Thanh Tran-Ngoc , Greg Angelo Gonzales Nonato , Hirotaka Sato This is my paper · ORCID
classification cs.RO
keywords insect-computerhybridrobotcyborgcockroachautomaticassemblyroboticarmmanipulationbipolarelectrodesdeeplearningsegmentationlocomotioncontrolmulti-agentcoverage
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 tries to prove that cyborg cockroaches can be mass-produced by replacing fragile, hour-long manual surgery with a 68-second automatic assembly line. The line uses a robotic arm that implants custom bipolar electrodes into a membrane between the pronotum and mesothorax, guided by a deep-learning camera that finds the correct insertion point. Five automatically assembled cockroaches steered and decelerated as well as manually assembled ones, and a team of four covered 80.25% of an obstructed outdoor square in 10 minutes 31 seconds. If the claim holds, insect-computer hybrids move from craft objects to factory products.

What carries the argument

The load-bearing object is a custom bipolar electrode: a 0.6 mm thick 3D-printed plastic microneedle with a hook, selectively plated with copper, that is small enough to puncture the soft intersegmental membrane yet anchored so it does not fall out. The other half of the machinery is the vision-guided manipulation chain: a fixation structure lifts the pronotum 1.9 mm to expose the membrane; a TransUNet segmentation model trained with DSC loss finds the reference point pR at the middle of the posterior pronotum edge; and a UR3e arm, after calibrating camera depth, implants the backpack's electrodes at a fixed pitch angle of 162.7 degrees, then presses the backpack's four branches onto the metathorax. The electrode's impedance below 70 ohms and the new pronotum stimulation site together make the strong turning and deceleration reactions possible.

What would settle it

Re-run the automatic assembly on a batch of new cockroaches with no human correction, then dissect each one to see whether both electrode microneedles actually punctured the intersegmental membrane; if the implant success rate falls, or if steering and deceleration responses on the new batch fail to match the reported five-animal results, the factory-scale claim fails.

Watch

Extended reading notes

Core claim

The central discovery is that precise steering and deceleration of a Madagascar hissing cockroach can be achieved by implanting hook-shaped bipolar electrodes into the intersegmental membrane between the pronotum and mesothorax, and that this implant can be done by a robot instead of a human without losing control quality. Automatically assembled robots turned 70.9 degrees left and 79.5 degrees right during 0.4 s stimulations, statistically indistinguishable from manual assembly ($P = 0.62$ and $P = 0.50$), and slowed from 6.3 cm/s to 2.0 cm/s under deceleration stimulation, also indistinguishable ($P = 0.21$). The assembly completes in 68 s, versus more than an hour for one manual preparation, and a multi-agent trial reached 80.25% coverage of a 2 by 2 meter obstructed outdoor terrain in 10 minutes 31 seconds. The paper also reports the first deceleration control for insect-computer hybrids, obtained by stimulating both outer electrodes and contracting both forelegs.

Load-bearing premise

The entire process depends on the segmentation model, trained on only nine cockroach images, finding the reference point pR on each new insect; if it misses by more than the electrode's 0.6 mm thickness, the electrodes land outside the soft membrane and control fails.

Editorial extensions

If this is right

  • Production time per hybrid drops from over an hour of manual surgery to 68 seconds, so a four-robot team can be assembled in 7 minutes 48 seconds.
  • Automatic assembly removes operator variability: the left-right turning-angle imbalance falls from 21.7% in manual assembly to 10.8%.
  • The pronotum stimulation protocol yields turns over 70 degrees in 0.4 s and a 68.2% speed reduction in 0.4 s, using 40% of the stimulation time and 75% of the voltage of an earlier non-invasive method.
  • Multiple automatically assembled hybrids can be deployed as a swarm, covering 80.25% of an obstructed 2 by 2 meter outdoor terrain in 10 minutes 31 seconds, compared with 14.00 to 45.75% for a single insect.

Reading between the lines

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

  • If the vision model is retrained on more than nine cockroach images, the same factory cell could likely hold its implant accuracy across wider natural size variation, and possibly transfer to other insects with a comparable pronotum-mesothorax membrane.
  • The reported 68 seconds is the robot's cycle time while a cockroach is already fixed and anesthetized; per-robot wall-clock throughput for a factory would additionally include anesthesia, recovery time, and the 4-hour rest used before deployment.
  • The same backpack form factor could carry other sensors, so the assembly line's value would increase if the reference-point detector is extended to locate additional body landmarks.
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 / 4 minor

Summary. The paper presents an automatic assembly pipeline for cyborg cockroaches. The system fixes an anesthetized cockroach in a custom structure, uses a TransUNet-based vision model to locate a reference point on the pronotum, and then commands a UR3e robotic arm to implant a custom bipolar electrode backpack into the exposed intersegmental membrane between pronotum and mesothorax. The authors report that one assembly takes 68 seconds, that automatically assembled robots exhibit steering and deceleration performance statistically indistinguishable from manually assembled ones, and that four such robots covered 80.25% of an obstructed 2 m x 2 m outdoor terrain in 10 minutes 31 seconds. The paper also introduces a pronotum stimulation protocol and a multi-material 3D-printed, electroless-plated bipolar electrode.

Significance. If the central claims hold, the work would be a meaningful step toward scalable production of insect-computer hybrid robots, which is an active and application-oriented area. The strengths include the custom electrode fabrication with FEA-based implantation analysis and ASTM-graded adhesion, the neural-recording-based selection of stimulation voltage, the inclusion of both steering and deceleration control, and a real multi-agent outdoor coverage demonstration. The manuscript is also refreshingly concrete about hardware details. However, the quantitative support for the assembly-time and equivalence claims is currently thin, and several load-bearing numbers are not tied to physical tolerances or attempt statistics. The core idea is plausible and the gaps appear addressable with additional measurements and reporting rather than being fatal.

major comments (4)
  1. [Sec. 4.3.2 and Table 2] The localization accuracy of the implantation reference point pR is reported only as an MSE of 1.695 pixels for the TransUNet model with DSC loss. The paper never converts this pixel error into physical units, so it cannot be checked against the electrode thickness of 0.6 mm and the membrane exposure of about 1.9 mm reported in Sec. 2.3.1. Since an erroneous pR would cause the robotic arm to miss the intersegmental membrane and invalidate all downstream control results, the authors should report the physical pixel-to-millimeter scale for the fixed camera and 256x256 crop, give the per-image pR error distribution, and state explicitly how many implantations fell within the membrane tolerance.
  2. [Sec. 2.3.3 / Sec. 2.4] No assembly success rate is reported. The text states that one assembly takes 68 seconds and that five automatically assembled robots were used for locomotion tests and four for the coverage mission, but it never states how many assembly attempts were made, how many failures occurred, or whether any failed assemblies required manual intervention. If the reported robots are the survivors of a larger pool, the claimed 68-s cycle time and the 'matching control' comparison do not represent the automatic process as deployed. The authors should report the total number of attempts, per-step failure rates, and the handling of failures, e.g., whether retries are included in the cycle time.
  3. [Sec. 2.3.3 vs. Sec. 2.4] There is an internal inconsistency in the assembly time figures. Sec. 2.3.3 states that 'the entire assembly process took 68 seconds,' while Sec. 2.4 says that four robots were assembled within 7 minutes 48 seconds, which is 117 seconds per robot. These numbers cannot both describe the same nominal cycle. The authors should reconcile them, for example by clarifying whether the 7 min 48 s includes batch setup, anesthetization, calibration, or idle time between robots, or by correcting one of the two figures. Because the central claim is that the automatic strategy reduces preparation time, this discrepancy must be resolved.
  4. [Sec. 2.4 and Fig. 5A] The claims of matching performance rely on Student's t-tests with P = 0.62, P = 0.50, and P = 0.21 for left turn, right turn, and deceleration, respectively, with N = 5 per group. Absence of a significant difference is not evidence of equivalence, especially at this sample size. The authors should report confidence intervals for the differences, provide an equivalence margin justified by application requirements, or use two one-sided tests (TOST). Without this, the statement that automatically assembled robots 'matched' manually assembled ones is not quantitatively established.
minor comments (4)
  1. [Sec. 2.1] The light intensity is given as '89 MW/cm2'; this is almost certainly a typo for 'mW/cm2' or 'mW/cm²'.
  2. [Sec. 4.2.1] In the finite element simulation description, a z-direction displacement of '50 mm' is stated for the microneedle structure. Given the electrode dimensions and the membrane exposure of about 1.9 mm, this value seems inconsistent; please clarify the actual implantation depth and boundary conditions.
  3. [Tables 1 and 2] The MSE values for pR are reported without units. Since they are computed in pixels, the caption should state 'MSE in pixels' and, ideally, also give the physical pixel size at the working distance.
  4. [Sec. 2.2] The claim that the new protocol 'increases the maximum steering speed by over five times' is stated without the exact comparison value from reference 7; reporting the precise ratio and the statistical basis would strengthen the comparison.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the 68 s assembly time, control comparisons, and coverage are measured outcomes, and the vision pR model is evaluated on held-out test images. Self-citations are used as benchmarks, not as load-bearing inputs.

full rationale

The paper's derivation chain is empirical. The stimulation voltage (3.0 V) is selected from measured neural spike counts at different voltages (Sec 4.4, Fig. 3A), not from an equation that assumes the resulting control behavior. The automatically assembled robots' steering angles, angular speeds, and deceleration are VICON-measured outcomes compared with manually assembled controls via t-tests (Sec 2.4, Fig. 5A); manual assembly is a control condition, not a fitted input or predicted quantity. The vision-guided localization of pR uses TransUNet trained on a small labeled set and evaluated on 20 held-out test images with reported mIoU, mDSC, and MSE (Sec 4.3.2, Tables 1-2); this is a standard held-out evaluation, not a fitted parameter renamed as a prediction. The 68 s assembly time is a timed observation (Sec 2.3.3), and the 80.25% coverage in 10 min 31 s is a measured multi-agent outcome (Sec 2.4). Self-citations (refs 7, 8, 13) appear as prior benchmarks for non-invasive electrodes, the methyl salicylate booster, and previous manual preparation time; none is invoked as a uniqueness theorem, none defines the present result, and none replaces a measurement. No equation in the paper defines one claimed output in terms of another claimed output, and no fitted parameter is relabeled as a prediction. The small training set and the unreported assembly-attempt success rate are legitimate correctness and robustness concerns, but they are not circularity.

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

The paper introduces no new natural entities (no new particles, forces, or conserved quantities). The bipolar electrode and backpack are engineered devices, not postulated entities with independent falsifiable handles. The central empirical results rest on a handful of tuned parameters (voltage, frequency, fixture distances, pitch angle) and on domain assumptions about cockroach anatomy and sensor accuracy.

free parameters (5)
  • Stimulation voltage = 3.0 V
    Chosen from the neural spike plateau in Fig. 3A (N=3 cockroaches); 3.0 V is the highest voltage before spike count declines at 4.0 V. All locomotion control depends on this value.
  • Stimulation frequency and duration = 42 Hz, 0.4 s
    Used for all locomotion tests; the paper does not justify these values or show a parameter sweep.
  • Fixation lowered distance d = 3.5 mm
    Chosen because pronotum lifting height h saturates at about 1.9 mm for d=3.5-4.0 mm (P=0.31); larger d risks unnecessary pressure on the insect.
  • Implantation pitch angle alpha = 162.7 degrees
    Midpoint of the measured lower threshold 157.8 +/- 1.5 deg and upper threshold 167.5 +/- 2.2 deg (N=5 insects); a single fixed angle is used for all assemblies.
  • Electroless plating time = 16 minutes
    Selected to balance plating thickness, conductivity, and parasitic capacitance based on the curve in Fig. 2G.
assumptions (5)
  • domain assumption Neural spike counts in the neck nerve cord are a monotonic proxy for stimulation effectiveness and behavioral response.
    Used in Sec 2.2 to select 3.0 V; no direct evidence links neck spike counts to the foreleg contraction or turning behavior.
  • domain assumption The intersegmental membrane between pronotum and mesothorax can be reliably exposed by lifting the pronotum with the 3D-printed structure, and the achieved clearance (h about 1.9 mm) is sufficient across individual cockroaches.
    Measured on 10 cockroaches; anatomical variability beyond this sample is not characterized.
  • ad hoc to paper The 29-image dataset (9 training images plus augmentation) adequately represents the size and shape distribution of the cockroach population for training a segmentation model.
    Sec 4.3.2; a 9-image training set is extraordinarily small and may not capture population variance, making the reported mIoU and MSE optimistic.
  • standard math Student's t-test assumptions (independence, approximate normality, equal variances) hold for the behavioral measurements.
    Sec 2.4 uses t-tests with N=5 per group; no normality or variance checks are reported.
  • domain assumption UWB localization error is negligible relative to the 10 cm grid used for coverage measurement.
    Sec 4.6; anchor setup and label placement are described, but no localization accuracy characterization is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cyborg Insect Factory: Automatic Assembly System to Build up Insect-computer Hybrid Robot Based on Vision-guided Robotic Arm Manipulation of Custom Bipolar Electrodes." pith.science (2026). https://pith.science/paper/VMGMMABS

@misc{pith2026241113164,
  author       = {Pith},
  title        = {Pith review of: Cyborg Insect Factory: Automatic Assembly System to Build up Insect-computer Hybrid Robot Based on Vision-guided Robotic Arm Manipulation of Custom Bipolar Electrodes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VMGMMABS}},
  note         = {Machine review of arXiv:2411.13164}
}
read the original abstract

The advancement of insect-computer hybrid robots holds significant promise for navigating complex terrains and enhancing robotics applications. This study introduced an automatic assembly method for insect-computer hybrid robots, which was accomplished by mounting backpack with precise implantation of custom-designed bipolar electrodes. We developed a stimulation protocol for the intersegmental membrane between pronotum and mesothorax of the Madagascar hissing cockroach, allowing for bipolar electrodes' automatic implantation using a robotic arm. The assembly process was integrated with a deep learning-based vision system to accurately identify the implantation site, and a dedicated structure to fix the insect (68 s for the whole assembly process). The automatically assembled hybrid robots demonstrated steering control (over 70 degrees for 0.4 s stimulation) and deceleration control (68.2% speed reduction for 0.4 s stimulation), matching the performance of manually assembled systems. Furthermore, a multi-agent system consisting of 4 hybrid robots successfully covered obstructed outdoor terrain (80.25% for 10 minutes 31 seconds), highlighting the feasibility of mass-producing these systems for practical applications. The proposed automatic assembly strategy reduced preparation time for the insect-computer hybrid robots while maintaining their precise control, laying a foundation for scalable production and deployment in real-world applications.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 42 canonical work pages

  1. [1]

    These benefits have driven the development both mechanically structured robots and insect-computer hybrid robots (i.e., biobots or cyborg insects)

    INTRODUCTION Various insect-scale robots have been developed offering significant advantages in maneuverability within complex and narrow terrains (1–5). These benefits have driven the development both mechanically structured robots and insect-computer hybrid robots (i.e., biobots or cyborg insects). Despite their similar size , insect-computer hybrid rob...

    work page

  2. [2]

    RESULTS AND DISCUSSION 2.1 Bipolar Electrode Due to the difficulty of fixing the antennae and abdomen and implanting electrodes using the robotic arm, this study focused solely on the pronotum. To induce the cockroach’s turning locomotion and minimize implantation sites , a pair of bipolar electrodes were designed and implanted on the left and right sides...

    work page

  3. [3]

    Locomotion studies and neural recordings on the hybrid robots reflected the reliability of the implantation and stimulation

    CONCLUSION This study developed an automatic assembly strategy for insect-computer hybrid robot based on a newly developed pronotum stimulation protocol. Locomotion studies and neural recordings on the hybrid robots reflected the reliability of the implantation and stimulation. Under this assembly procedure, the hybrid robots could be prepared in a short ...

    work page

  4. [4]

    These insects were commonly used for the hybrid robots (7,15), which had mature stimulation protocol to compare with our new one

    MATERIALS AND METHODS 4.1 Insect Platform Male Madagascar hissing cockroaches (5-6 cm) were used for this study . These insects were commonly used for the hybrid robots (7,15), which had mature stimulation protocol to compare with our new one. All the cockroaches used were supplied carrots and water weekly. The cockroaches used for hybrid robot assembly w...

    work page

  5. [5]

    Multidirectional Planar Motion Transmission on a Single-Motor Actuated Robot via Microscopic Galumphing

    Tang L, Wang C, Ma S, Li Y , Li B. Multidirectional Planar Motion Transmission on a Single-Motor Actuated Robot via Microscopic Galumphing. Adv Sci. 2024;11(9):2307738

    work page 2024

  6. [6]

    A biologically inspired, flapping-wing, hybrid aerial-aquatic microrobot

    Chen Y , Wang H, Helbling EF, Jafferis NT, Zufferey R, Ong A, et al. A biologically inspired, flapping-wing, hybrid aerial-aquatic microrobot. Sci Robot. 2017 Oct 25;2(11):eaao5619

    work page 2017

  7. [7]

    Sideways Walking Control of a Cyborg Beetle

    Nguyen HD, Tan PZ, Sato H, V o-Doan TT. Sideways Walking Control of a Cyborg Beetle. IEEE Trans Med Robot Bionics. 2020 Aug;2(3):331–7

    work page 2020

  8. [8]

    Intelligent Insect–Computer Hybrid Robot: Installing Innate Obstacle Negotiation and Onboard Human Detection onto Cyborg Insect

    Tran-Ngoc PT, Le DL, Chong BS, Nguyen HD, Dung VT, Cao F, et al. Intelligent Insect–Computer Hybrid Robot: Installing Innate Obstacle Negotiation and Onboard Human Detection onto Cyborg Insect. Adv Intell Syst. 2023 Jan 27;n/a(n/a):2200319

    work page 2023

Show all 49 references
  1. [9]

    A wireless controlled robotic insect with ultrafast untethered running speeds

    Liu Z, Zhan W, Liu X, Zhu Y , Qi M, Leng J, et al. A wireless controlled robotic insect with ultrafast untethered running speeds. Nat Commun. 2024 May 8;15(1):3815

    2024

  2. [10]

    The long-acting walking control of a cockroach bio-bot for vision-aided pipeline navigation

    Ma S, Liu S, Yang S, Chen Y , Li Y , Li B. The long-acting walking control of a cockroach bio-bot for vision-aided pipeline navigation. In: 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). 2023. p. 1–4

    2023

  3. [11]

    Integration of body-mounted ultrasoft organic solar cell on cyborg insects with intact mobility

    Kakei Y , Katayama S, Lee S, Takakuwa M, Furusawa K, Umezu S, et al. Integration of body-mounted ultrasoft organic solar cell on cyborg insects with intact mobility. Npj Flex Electron. 2022 Sep 5;6(1):1–9

    2022

  4. [12]

    Resilient conductive membrane synthesized by in-situ polymerisation for wearable non-invasive 27 electronics on moving appendages of cyborg insect

    Lin Q, Li R, Zhang F, Kai K, Ong ZC, Chen X, et al. Resilient conductive membrane synthesized by in-situ polymerisation for wearable non-invasive 27 electronics on moving appendages of cyborg insect. Npj Flex Electron. 2023 Sep 1;7(1):1–10

    2023

  5. [13]

    A newly developed chemical locomotory booster for cyborg insect to sustain its activity and to enhance covering performance

    Lin Q, Kai K, Nguyen HD, Sato H. A newly developed chemical locomotory booster for cyborg insect to sustain its activity and to enhance covering performance. Sens Actuators B Chem. 2024 Jan 15;399:134774

    2024

  6. [14]

    CameraRoach: various electronic backs packs for Search and Rescue

    Rasakatla S, Suzuki T, Tenma W, Mizuuchi I, Indurkhya B. CameraRoach: various electronic backs packs for Search and Rescue. In: 2021 IEEE International Conference on Robotics and Biomimetics (ROBIO). 2021. p. 1300– 3

    2021

  7. [15]

    Effective Stimulus Parameters for Directed Locomotion in Madagascar Hissing Cockroach Biobot

    Erickson JC, Herrera M, Bustamante M, Shingiro A, Bowen T. Effective Stimulus Parameters for Directed Locomotion in Madagascar Hissing Cockroach Biobot. PLOS ONE. 2015 Aug 26;10(8):e0134348

    2015

  8. [16]

    A Navigation Algorithm to Enable Sustainable Control of Insect-Computer Hybrid Robot with Stimulus Signal Regulator and Habituation-Breaking Function

    Li R, Lin Q, Kai K, Nguyen HD, Sato H. A Navigation Algorithm to Enable Sustainable Control of Insect-Computer Hybrid Robot with Stimulus Signal Regulator and Habituation-Breaking Function. Soft Robot. 2023 Dec 29

    2023

  9. [17]

    Movement Optimization for a Cyborg Cockroach in a Bounded Space Incorporating Machine Learning

    Ariyanto M, Refat CMM, Hirao K, Morishima K. Movement Optimization for a Cyborg Cockroach in a Bounded Space Incorporating Machine Learning. Cyborg Bionic Syst. 2023 Jan;4:0012

    2023

  10. [18]

    Natural-artificial hybrid swarm: Cyborg-insect group navigation in unknown obstructed soft terrain

    Bai Y , Ngoc PTT, Nguyen HD, Le DL, Ha QH, Kai K, et al. Natural-artificial hybrid swarm: Cyborg-insect group navigation in unknown obstructed soft terrain. http://arxiv.org/abs/2403.17392 (2024)

    2024 arXiv

  11. [19]

    Radio-Controlled Cyborg Beetles: A Radio-Frequency System for Insect Neural Flight Control

    Sato H, Peeri Y , Baghoomian E, Berry CW, Maharbiz MM. Radio-Controlled Cyborg Beetles: A Radio-Frequency System for Insect Neural Flight Control. In: 2009 IEEE 22nd International Conference on Micro Electro Mechanical Systems

    2009

  12. [20]

    The Autonomous Pipeline Navigation of a Cockroach Bio-Robot with Enhanced Walking Stimuli

    Ma S, Chen Y , Yang S, Liu S, Tang L, Li B, et al. The Autonomous Pipeline Navigation of a Cockroach Bio-Robot with Enhanced Walking Stimuli. Cyborg Bionic Syst. 2023 Nov 8;4:0067

    2023

  13. [21]

    Feedback Control-Based Navigation of a Flying Insect- Machine Hybrid Robot

    Li Y , Wu J, Sato H. Feedback Control-Based Navigation of a Flying Insect- Machine Hybrid Robot. Soft Robot. 2018 Aug 28;5(4):365–74

    2018

  14. [22]

    Launching of a Cyborg Locust via Co-Contraction Control of Hindleg Muscles

    Ma S, Liu P, Liu S, Li Y , Li B. Launching of a Cyborg Locust via Co-Contraction Control of Hindleg Muscles. IEEE Trans Robot. 2022;1–12

    2022

  15. [23]

    Minimal navigation solution for a swarm of tiny flying robots to explore an unknown environment

    McGuire KN, De Wagter C, Tuyls K, Kappen HJ, de Croon GCHE. Minimal navigation solution for a swarm of tiny flying robots to explore an unknown environment. Sci Robot. 2019 Oct 23;4(35):eaaw9710

    2019

  16. [24]

    Predictive control of aerial swarms in cluttered environments

    Soria E, Schiano F, Floreano D. Predictive control of aerial swarms in cluttered environments. Nat Mach Intell. 2021 Jun;3(6):545–54. 28

    2021

  17. [25]

    Collision avoidance by running insects: antennal guidance in cockroaches

    Baba Y , Tsukada A, Comer CM. Collision avoidance by running insects: antennal guidance in cockroaches. J Exp Biol. 2010 Jul 1;213(13):2294–302

    2010

  18. [26]

    Contact Reduction with Bounded Stiffness for Robust Sim- to-Real Transfer of Robot Assembly

    Vuong N, Pham QC. Contact Reduction with Bounded Stiffness for Robust Sim- to-Real Transfer of Robot Assembly. In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 2023. p. 361–7

    2023

  19. [27]

    An Ultralightweight and Living Legged Robot

    V o Doan TT, Tan MYW, Bui XH, Sato H. An Ultralightweight and Living Legged Robot. Soft Robot. 2018 Feb;5(1):17–23

    2018

  20. [28]

    Mechanosensation and Adaptive Motor Control in Insects

    Tuthill JC, Wilson RI. Mechanosensation and Adaptive Motor Control in Insects. Curr Biol. 2016 Oct 24;26(20):R1022–38

    2016

  21. [29]

    (Fig. 2B). Initially, multi-material DLP3D printing was used to create the precursor structure of the bipolar electrodes, composed of a normal resin and an active precursor. The active precursor contains a catalytic factor that facilitates selective metal deposition during ele...

  22. [30]

    Deciding Which Way to Go: How Do Insects Alter Movements to Negotiate Barriers? Front Neurosci

    Ritzmann RE, Harley CM, Daltorio KA, Tietz BR, Pollack AJ, Bender JA, et al. Deciding Which Way to Go: How Do Insects Alter Movements to Negotiate Barriers? Front Neurosci. 2012 Jul 6

    2012

  23. [31]

    Characterization of obstacle negotiation behaviors in the cockroach, Blaberus discoidalis

    Harley CM, English BA, Ritzmann RE. Characterization of obstacle negotiation behaviors in the cockroach, Blaberus discoidalis. J Exp Biol. 2009 May 15;212(10):1463–76

    2009

  24. [32]

    New Metal–Plastic Hybrid Additive Manufacturing for Precise Fabrication of Arbitrary Metal Patterns on External and Even Internal Surfaces of 3D Plastic Structures

    Song K, Cui Y , Tao T, Meng X, Sone M, Yoshino M, et al. New Metal–Plastic Hybrid Additive Manufacturing for Precise Fabrication of Arbitrary Metal Patterns on External and Even Internal Surfaces of 3D Plastic Structures. ACS Appl Mater Interfaces. 2022 Oct 19;14(41):46896–911

    2022

  25. [33]

    Modified polymer 3D printing enables the formation of functionalized micro-metallic architectures

    Perera ATK, Wu K, Wan WY , Song K, Meng X, Umezu S, et al. Modified polymer 3D printing enables the formation of functionalized micro-metallic architectures. Addit Manuf. 2023 Jan 5;61:103317

    2023

  26. [34]

    3D Programming of Patterned Heterogeneous Interface for 4D Smart Robotics

    Song K, Xiong C, Zhang Z, Wu K, Wan W, Wang Y , et al. 3D Programming of Patterned Heterogeneous Interface for 4D Smart Robotics. http://arxiv.org/abs/2312.14511 (2023)

    2023 arXiv

  27. [35]

    Development of hybrid composites reinforced with biocarbon/carbon fiber system

    Andrzejewski J, Mohanty AK, Misra M. Development of hybrid composites reinforced with biocarbon/carbon fiber system. The comparative study for PC, ABS and PC/ABS based materials. Compos Part B Eng. 2020 Nov 1;200:108319

    2020

  28. [36]

    The comparative investigation of the stick insect and cockroach models in the study of insect locomotion

    Ayali A, Borgmann A, Büschges A, Couzin-Fuchs E, Daun-Gruhn S, Holmes P. The comparative investigation of the stick insect and cockroach models in the study of insect locomotion. Curr Opin Insect Sci. 2015 Dec 1;12:1–10

    2015

  29. [37]

    Smart insect-computer hybrid robots empowered with enhanced obstacle avoidance capabilities using onboard monocular camera

    Li R, Lin Q, Tran-Ngoc PT, Le DL, Sato H. Smart insect-computer hybrid robots empowered with enhanced obstacle avoidance capabilities using onboard monocular camera. Npj Robot. 2024 Jun 24;2(1):1–10

    2024

  30. [38]

    Centerline Boundary Dice Loss for Vascular Segmentation

    Shi P, Hu J, Yang Y , Gao Z, Liu W, Ma T. Centerline Boundary Dice Loss for Vascular Segmentation. http://arxiv.org/abs/2407.01517 (2024) 29

    2024 arXiv

  31. [39]

    Adaptive end-effector pose control for tomato harvesting robots

    Wang D, Dong Y , Lian J, Gu D. Adaptive end-effector pose control for tomato harvesting robots. J Field Robot. 2023;40(3):535–51

    2023

  32. [40]

    Biobotic insect swarm based sensor networks for search and rescue

    Bozkurt A, Lobaton E, Sichitiu M, Hedrick T, Latif T, Dirafzoon A, et al. Biobotic insect swarm based sensor networks for search and rescue. In: Signal Processing, Sensor/Information Fusion, and Target Recognition XXIII. SPIE; 2014. p. 498– 503

    2014

  33. [41]

    A Survey on Vision Transformer

    Han K, Wang Y , Chen H, Chen X, Guo J, Liu Z, et al. A Survey on Vision Transformer. IEEE Trans Pattern Anal Mach Intell. 2023 Jan;45(1):87–110

    2023

  34. [42]

    Segment Anything

    Kirillov A, Mintun E, Ravi N, Mao H, Rolland C, Gustafson L, et al. Segment Anything. http://arxiv.org/abs/2304.02643 (2023)

    2023 arXiv

  35. [43]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. http://arxiv.org/abs/2010.11929 (2021)

    2021 arXiv

  36. [44]

    How to Train Vision Transformer on Small-scale Datasets? [Internet]

    Gani H, Naseer M, Yaqub M. How to Train Vision Transformer on Small-scale Datasets? [Internet]. http://arxiv.org/abs/2210.07240 (2022)

    2022 arXiv

  37. [45]

    U-Net: Convolutional Networks for Biomedical Image Segmentation

    Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. http://arxiv.org/abs/1505.04597 (2015)

    2015 arXiv

  38. [46]

    TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation

    Chen J, Lu Y , Yu Q, Luo X, Adeli E, Wang Y , et al. TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation. http://arxiv.org/abs/2102.04306 (2021)

    2021 arXiv

  39. [47]

    Rethinking Atrous Convolution for Semantic Image Segmentation

    Chen LC, Papandreou G, Schroff F, Adam H. Rethinking Atrous Convolution for Semantic Image Segmentation. http://arxiv.org/abs/1706.05587 (2017)

    2017 arXiv

  40. [48]

    Deep Residual Learning for Image Recognition [Internet]

    He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition [Internet]. http://arxiv.org/abs/1512.03385 (2015)

    2015 arXiv

  41. [49]

    ImageNet: A large-scale hierarchical image database

    Deng J, Dong W, Socher R, Li LJ, Li K, Fei-Fei L. ImageNet: A large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition. 2009. p. 248–55. Acknowledgments: The authors thank Mr. See To Yu Xiang, Dr. Kazuki Kai, Dr Duc Long Le, ...

    2009

Pith tools

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