{"id":"308d73b4-c718-4edf-bf76-2bfb38645ae0","arxiv_id":"2505.00969","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"An autonomous tape-application end-device induces surface wrinkles in an inflating vine robot, producing consistent 21-degree planar turns in real time, though consecutive-turn stability remains limited.","lead":"A growing soft robot now steers itself in real time by automatically taping wrinkles into its body as it grows, making repeated 21-degree turns on a flat surface. The system is a step toward practical steerable vine robots, but consecutive turns degrade after about ten wrinkles and the path planner was not tested in the reported experiments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The planner-based 'arbitrary many turns' claim is unsupported: Algorithm 1 is never run, and same-direction sequences degrade and fail after about ten wrinkles.","rationale":"Read in good faith, the paper's hardware contribution—real-time wrinkle induction with a two-tape compensation mechanism producing repeatable ~21° turns—is supported by the single-turn and short alternating-sequence data. However, the abstract claims 'arbitrary many turns' and 'repeated 21-degree turns using a Dubins path planner with minimal error.' Neither is established. The algorithm in Section V is only a sketch and returns a single intersection rather than a turn sequence; the experimental section contains no planner execution. The same-direction stress test that is reported fails after ten turns, which is a direct contradiction of 'arbitrary many turns' and is acknowledged in Section VII-B. This is the most load-bearing concern because if the intended contribution is the complete system including path following, the missing experiment is decisive; if the intended contribution is only the single-turn mechanism, then the paper should be reframed to avoid the planner claim. The reader identified the same evidence but selected wrinkle-length constancy as the weakest assumption; I partially agree, since D/tape alignment is a secondary mechanism behind the observed same-direction degradation. A single planner-execution experiment with a long same-direction arc would settle whether the broad claim holds. Because the hardware result and short-sequence control are still credible, the appropriate verdict remains CONDITIONAL, matching the reader.","tokens_in":6780,"tokens_out":8922,"duration_ms":99243,"concrete_test":"Run a single end-to-end experiment in which Algorithm 1 is used to plan a path with at least ten consecutive same-direction turns (e.g., a 210° right arc) and the physical robot executes it under motion capture. Record per-wrinkle angle and final pose error. If the robot fails before completing the path, or if final orientation error exceeds the 21±2° per-turn tolerance accumulated over the sequence, the 'arbitrary many turns using a Dubins path planner' claim fails. As a control, repeat the same path with alternating left-right turns to verify that the degradation is specific to sustained same-direction turning.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single-turn fixed-angle result (21.5±1.5°, alternating 21±2°) is credible as a hardware demonstration. The load-bearing gap is the extension to 'arbitrary many turns using a Dubins path planner.' Section V's Algorithm 1 returns a single intersection point and path length; it never generates a sequence of discrete 21° turns, and no experiment in Section VI executes it. More seriously, the only multi-turn same-direction data contradicts the broad claim: Section VI-A reports that beyond ten consecutive wrinkles the bend angle systematically decreases and the system fails, and Section VI-B attributes this to tape drift and air leakage. Since Dubins paths can require sustained same-direction arcs, a planner that outputs such arcs would hit exactly the regime where the hardware is known to fail. Thus the central contribution as stated in the abstract is not merely unvalidated; the available data indicate a hard limit well short of 'arbitrary many turns.' The authors' own Limitations section acknowledges fixed-angle, planar-only, and material-stability limits, but the abstract's path-planner claim remains unsupported by any experiment.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a vine robot steering system that creates controlled turns in real time by using an end-device to induce surface wrinkles in the growing tube and fix them with adhesive tape. The authors derive a geometric model that predicts a fixed turning angle of about 21 degrees, describe a modified Dubins path planner intended to convert sequences of these discrete turns into a navigable path, and report experiments measuring single-turn, consecutive same-direction, and alternating left-right turn performance. The central hardware claim is that single turns are repeatable at 21.5 degrees plus or minus 1.5 degrees, with alternating sequences averaging 21 degrees plus or minus 2 degrees over the first ten turns.","tokens_in":6945,"tokens_out":3699,"duration_ms":40337,"significance":"If the single-turn result is robust, the contribution is a genuinely simple, real-time steering mechanism for a soft growing robot that preserves the soft body and does not require pre-shaped paths or rigid actuators. The external material feed and compensation mechanism are a plausible engineering advance, and the paper provides a useful geometric model for estimating the turn angle from wrinkle geometry, subject to the concerns below. However, the broader claims in the abstract and conclusion about 'arbitrary many turns' and a working Dubins path planner are not substantiated by the experiments, and the available data indicate a hard limit of roughly ten consecutive same-direction turns. The paper would be significantly strengthened by reframing these claims and by providing the missing experimental and algorithmic details.","major_comments":[{"comment":"The model uses D=19 mm in Eq. (1), but Section III-A states that the wrinkle inducer arm's 15-degree rotation produces a precise 18 mm fold. This discrepancy is unexplained. Because the predicted angle scales linearly with D, the agreement between the model's 21.44 degrees and the experimental 21.5 degrees depends on the choice of 19 mm. Please justify the 19 mm value (for example, by reporting a direct measurement of the fold length) or redo the prediction with the design value D=18 mm, which yields theta' = 20.3 degrees and still lies within the reported experimental uncertainty. As written, the model appears to use an adjusted parameter rather than a first-principles value.","section":"Section IV, Eq. (1)"},{"comment":"The abstract's claim of 'arbitrary many turns using a Dubins path planner' is contradicted by the paper's own results. Algorithm 1 is never executed in any experiment in Section VI, and Section VI-A reports that beyond ten consecutive same-direction wrinkles the bend angle systematically decreases and the system fails. This directly refutes the 'arbitrary many turns' wording. The abstract and conclusion should be revised to state that the planner is proposed but not experimentally validated, and that sustained same-direction turning is currently limited to about ten wrinkles.","section":"Abstract, Section V, Section VI-A"},{"comment":"The pseudocode of Algorithm 1 does not implement the behavior described in the text. The text says that a 63-degree turn uses three consecutive 21-degree turns, but the algorithm as written searches for a single intersection point between discretized rays and returns a path length, not a sequence of discrete turn commands. There is no loop that would generate a series of 21-degree waypoints. Please provide the actual planning algorithm that generates discrete-turn sequences, or clearly label the current pseudocode as a conceptual sketch that has not been implemented.","section":"Section V, Algorithm 1"},{"comment":"The experimental section does not report a measurement protocol or the number of trials underlying the statistics in Fig. 7 and the text. There is no description of how the turning angle was measured (e.g., protractor, image processing, tracking markers), how many runs were performed for each scenario, or how the error bars are defined. Without this information, the claims of '21.5 plus or minus 1.5 degrees' and '21 plus or minus 2 degrees' cannot be independently assessed, and the paper's key quantitative result lacks adequate support. Please add a clear experimental methodology subsection.","section":"Section VI"}],"minor_comments":[{"comment":"The text states that the fold length is determined by the length of the support arms and later says that a 15-degree arm rotation produces an 18 mm fold. Please clarify which geometric quantity (arm length or rotation angle) is the controlled parameter and how the 18 mm value is obtained.","section":"Section III-A"},{"comment":"The sentence 'the outer arc of the wrinkle maintains this 2D length' is difficult to follow. Please expand the geometric derivation so that the factor of 2D in the length reduction, and hence theta = D/r, is clearly explained.","section":"Section IV"},{"comment":"The error analysis says that tape misalignment causes angular errors despite maintaining consistent wrinkle lengths, but this seems to imply that D remains constant while the angle changes. Please explain how D can stay fixed when the tape position relative to the wrinkle drifts, given that the model in Eq. (1) assumes D is the sole geometric determinant of the angle.","section":"Section VI-B"},{"comment":"The caption for Fig. 5 shows patterns R, RRR, and RLRLL but no quantitative angles or path measurements for these multi-turn demonstrations. Reporting the measured angles for each turn in these sequences would strengthen the evidence for multi-turn capability.","section":"Fig. 5"},{"comment":"The URL in reference [13] appears to contain spacing and possible transcription errors. Please verify that the link is complete and functional.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The hardware demonstration of repeatable single-turn steering is a solid engineering contribution that could be appropriate for a journal after revision. The main barrier is the mismatch between the claims in the abstract (arbitrary many turns, validated Dubins planner) and the evidence in the body, which shows a ten-turn limit and no planner experiments. This is fixable by scaling back the claims, adding the missing methodology, and resolving the D inconsistency. I would not recommend acceptance in the current form, but I see a clear path to a publishable paper if the authors address the four major comments."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real innovation here is the two-tape compensation mechanism: applying removable adhesive tape at the growing tip to create an asymmetric wrinkle, with the compensation step preserving that asymmetry. That is a genuine step beyond the pre-taped [9] and heat-welded permanent [11] approaches, and it is the paper's main contribution. The hardware demo backs up the core single-turn claim: 21.5° ± 1.5° for isolated turns, and 21° ± 2° for alternating sequences. The CAD files are posted, the mechanism is clearly described, and the authors are upfront about limitations. Credit where it is due: this is a working prototype with repeatable fixed-angle steering, which is more than many soft-robotics papers deliver.\n\nThe soft spots are real but mostly in the overreach, not in the core mechanism. The angle model uses D = 19 mm when the design section says the wrinkle inducer produces an 18 mm fold. That matters because D is the input to the geometric prediction; without an independent calibration of D, the 21.44° number is a consistency check, not a first-principles derivation. The paper should explain the 18-to-19 discrepancy and ideally measure D directly.\n\nThe broader \"arbitrary many turns\" claim is the load-bearing weakness. Algorithm 1 is never run; it returns a single intersection point and path length, not a sequence of discrete 21° turns. No experiment executes the planner. Worse, the only same-direction multi-turn data contradict the abstract: beyond ten consecutive wrinkles the bend angle systematically drops and the system fails, with the authors attributing this to tape drift and air leakage. Dubins paths can require long same-direction arcs, so a planner that outputs those arcs would hit exactly the regime where the hardware is known to degrade. The alternating-turn data are better, but only cover ten turns. So the paper supports repeated fixed-angle planar turns in the short term, not \"arbitrary many turns.\"\n\nOn citation pattern: the related work is appropriate for the subfield, though the references are somewhat thin on recent steering work beyond the immediate Hawkes/Satake lineage. That is a minor issue.\n\nNet: the single-turn result is credible and the mechanism is worth knowing about. The paper needs a tempering of the abstract, an explanation of D, and an actual planner experiment (or the planner claim removed) before publication. For a serious referee, this should go to review, not desk reject, but it needs revision. I would bring it to a reading group if the topic came up, and I would cite it for the two-tape mechanism if I worked on vine-robot steering.","headline":"Real-time tip-applied tape steering is a genuine mechanical contribution with a credible single-turn result, but the paper's 'arbitrary many turns' and Dubins claims are unsupported and its angle model has an unexplained parameter shift.","tokens_in":7504,"tokens_out":1267,"would_cite":true,"duration_ms":14312,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper shows a vine-style growing robot can be steered in real time by inducing surface wrinkles and fixing them with tape, producing repeatable ~21-degree turns.","keywords":["vine robot","soft robotics","real-time steering","surface wrinkles","adhesive tape","fixed-angle turns","Dubins path planning","growing robot"],"falsifier":"Measure the wrinkle fold length and tape lateral position on every wrinkle of a ten-turn sequence and compare each measured turn angle with θ′ = (Dπ/L) cos(45°); if the angle drifts while D and tape alignment stay fixed, the geometric model is wrong, and if D or tape alignment drifts first, the fixed-angle claim fails.","tokens_in":6544,"feed_emoji":"🤖","tokens_out":5847,"duration_ms":59269,"temperature":0.7,"pith_summary":"This paper claims that a vine-like growing soft robot can be steered in real time by mechanically creating a fold, or wrinkle, in the uninflated tube at the growth point and fixing that fold in place with adhesive tape. The design feeds material from outside the inflated tube so wrinkles can be induced before inflation, and a compensation mechanism makes the fold asymmetric so the tube bends consistently when it inflates. The model predicts each wrinkle turns the robot about 21 degrees on a flat surface, and experiments report 21.5 ± 1.5 degrees for single turns and 21 ± 2 degrees for alternating turns. A discrete shortest-path planner turns the fixed-angle constraint into a sequence of left/right 21-degree increments, so the robot can follow planar paths without internal rigid actuators.","feed_headline":"Tape-induced wrinkles steer soft vine robots in 21-degree turns","feed_subtitle":"On-demand left/right turns at 21 degrees per wrinkle were demonstrated; repeated alternating turns stayed within 2 degrees.","key_machinery":"The load-bearing mechanism is the wrinkle inducer and compensation system: two support arms lift the uninflated tube to create a fold while a lower bonding arm presses one of two exterior tape ribbons onto the fold, and a compensation arm first adheres the opposite ribbon so that side unfolds on inflation. The central identity is θ′ = (Dπ/L) cos(45°), the planar projection of the wrinkle-induced bend, where D is wrinkle length, L is flattened tube width, and 45° is the tape's quarter-point position around the circumference. This identity converts a fixed mechanical stroke into a repeatable turning angle and lets the controller treat each wrinkle as a discrete 21-degree turn.","core_discovery":"The central claim is that a Vine robot can be given arbitrary many planar turns in real time by inducing surface wrinkles at its growing end and taping them, without sacrificing the robot's soft body. Each wrinkle shortens one side of the tube by twice the wrinkle length, producing a turning angle θ = D/r = Dπ/L, which for D = 19 mm and L = 105 mm is 30.3 degrees in 3D; because the tape is placed at quarter-point positions, the planar projection is θ′ = θ cos(45°) ≈ 21.44°. Experiments confirm this: first-turn angle 21.5 ± 1.5°, alternating left/right sequences average 21 ± 2°, though ten consecutive same-direction turns degrade as material shifts and air leaks. The claim is that this mechanism is the first to combine real-time steering, multi-directional turns, and structural softness in a growing vine robot, with path planning reduced to discrete 21-degree steps.","pith_inferences":["The cos(45°) projection suggests a tunable steering law: if the tape's circumferential position could be varied, the same wrinkle mechanism would produce a continuous family of planar angles between 0° and 30.3°.","The degradation after consecutive same-direction turns points to a control fix the paper does not test: alternating turn directions, or periodically re-centering the tape, may act as an error-correcting strategy and extend the usable turn count.","The compensation principle—pre-adhering the opposite side so a symmetric fold unfolds asymmetrically—could transfer to other soft growing robots as a way to create programmed bends without internal actuators."],"forward_implications":["Because each wrinkle yields a repeatable fixed-angle turn, the robot can execute arbitrary planar trajectories as sequences of left/right 21-degree increments, with a shortest-path planner choosing the sequence.","The external material feed lets the wrinkle be formed before inflation, so turning control happens in real time at the growth point rather than being fixed before deployment.","Alternating left and right turns stay within 2 degrees for at least ten turns, so bidirectional steering is usable immediately; the robot remains a soft continuum structure throughout.","The fixed-angle limit is a design choice, not a physical ceiling: varying the wrinkle length would produce arbitrary angles, extending the same mechanism to general paths."],"supporting_citations":[{"why":"Supplies the everting vine-robot design and tape-configuration model that this external-feed, wrinkle-based system builds on.","marker":"[2]"},{"why":"Establishes the pre-taped asymmetric-constraint steering method that the paper automates in real time.","marker":"[9]"},{"why":"Shows a head-applied heat-welding bend that is permanent, motivating the removable-tape wrinkle fixation here.","marker":"[11]"},{"why":"Describes tip-extension growth with internal everting feed, the contrast case for the external feed used here.","marker":"[12]"},{"why":"Provides the shortest-path-with-curvature-constraint curves that the discrete 21-degree planner modifies.","marker":"[14]"}],"fun_headline_variants":["Tape wrinkles steer soft vine robots in 21° turns","Real-time 21° steering for vine robots via tape wrinkles","Vine robot turns: tape-induced wrinkles give 21° steps","On-demand 21° turns for vine robots using tape wrinkles","Soft robot steering: tape wrinkles enable 21° vine turns"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 21-degree result depends on the wrinkle length staying at 19 mm and the tape staying bonded at the quarter-point positions, and the paper's own error analysis shows these are precisely what drift after repeated turns.","fun_headline_variants_meta":{"raw":{"variants":["Tape wrinkles steer soft vine robots in 21° turns","Real-time 21° steering for vine robots via tape wrinkles","Vine robot turns: tape-induced wrinkles give 21° steps","On-demand 21° turns for vine robots using tape wrinkles","Soft robot steering: tape wrinkles enable 21° vine turns"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000308,"raw_usage":{"total_tokens":1712,"prompt_tokens":849,"completion_tokens":863,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":465,"completion_tokens_details":{"reasoning_tokens":776}},"tokens_in":465,"tokens_out":863,"duration_ms":8838,"temperature":1.0,"reasoning_tokens":776,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:29:57.113301+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the wrinkle fold length and tape lateral position on every wrinkle of a ten-turn sequence and compare each measured turn angle with θ′ = (Dπ/L) cos(45°); if the angle drifts while D and tape alignment stay fixed, the geometric model is wrong, and if D or tape alignment drifts first, the fixed-angle claim fails.","supporting_citations":[{"cited_title":"Design, modeling, control, and application of everting vine robots,","cited_arxiv_id":null,"evidence_quote":"Supplies the everting vine-robot design and tape-configuration model that this external-feed, wrinkle-based system builds on."},{"cited_title":"Novel growing robot with in- flatable structure and heat-welding rotation mechanism,","cited_arxiv_id":null,"evidence_quote":"Shows a head-applied heat-welding bend that is permanent, motivating the removable-tape wrinkle fixation here."},{"cited_title":"A soft, steerable continuum robot that grows via tip extension,","cited_arxiv_id":null,"evidence_quote":"Describes tip-extension growth with internal everting feed, the contrast case for the external feed used here."},{"cited_title":"On curves of minimal length with a constraint on average curvature, and with prescribed initial and terminal positions and tangents,","cited_arxiv_id":null,"evidence_quote":"Provides the shortest-path-with-curvature-constraint curves that the discrete 21-degree planner modifies."}],"review_version":1}