OSCAR: An Ovipositor-Inspired Self-Propelling Capsule Robot for Colonoscopy

Mostafa A. Atalla; Anand S. Sekar; Remi van Starkenburg; David J. Jager; Aimée Sakes; Michaël Wiertlewski; Paul Breedveld

cs.ROFeb 17, 2026

OSCAR: An Ovipositor-Inspired Self-Propelling Capsule Robot for Colonoscopy

Authors: Mostafa A. Atalla, Anand S. Sekar, Remi van Starkenburg, David J. Jager, Aimée Sakes, Michaël Wiertlewski, Paul Breedveld

Organizations: M. A. Atalla, A. S. Sekar, R. van Starkenburg, D. J. Jager, A. Sakes, and P. Breedveld are with the Department of BioMechanical Engineering, Faculty of Mechanical Engineering, Delft University of Technology, 2628 CD Delft, The Netherlands. · M. A. Atalla and M. Wiertlewski are with the Department of Cognitive Robotics, Faculty of Mechanical Engineering, Delft University of Technology, 2628 CD Delft, The Netherlands.

Abstract

Self-propelling robotic capsules eliminate shaft looping of conventional colonoscopy, reducing patient discomfort. However, reliably moving within the slippery, viscoelastic environment of the colon remains a significant challenge. We present OSCAR, an ovipositor-inspired self-propelling capsule robot that translates the transport strategy of parasitic wasps into a propulsion mechanism for colonoscopy. OSCAR mechanically encodes the ovipositor-inspired motion pattern through a spring-loaded cam system that drives twelve circumferential sliders in a coordinated, phase-shifted sequence. By tuning the motion profile to maximize the retract phase relative to the advance phase, the capsule creates a controlled friction anisotropy at the interface that generates net forward thrust. We developed an analytical model incorporating a Kelvin-Voigt formulation to capture the viscoelastic stick--slip interactions between the sliders and the tissue, linking the asymmetry between advance and retract phase durations to mean thrust, and slider-reversal synchronization to thrust stability. Comprehensive force characterization experiments in ex-vivo porcine colon revealed a mean steady-state traction force of 0.85 N, closely matching the model. Furthermore, experiments confirmed that thrust generation is speed-independent and scales linearly with the phase asymmetry, in agreement with theoretical predictions, underscoring the capsule's predictable performance and scalability. In locomotion validation experiments, OSCAR demonstrated robust performance, achieving an average speed of 3.08 mm/s, a velocity sufficient to match the cecal intubation times of conventional colonoscopy. By coupling phase-encoded friction anisotropy with a predictive model, OSCAR delivers controllable thrust generation at low normal loads, enabling safer and more robust self-propelling locomotion for robotic capsule colonoscopy.

Figures & tables

Fig. 1: Ovipositor-Inspired Capsule Robot for Colonoscopy (OSCAR). (a) Parasitic wasps use reciprocating sliding valves that alternately anchor and release to transport eggs. (b) OSCAR translates this principle into a compact capsule with reciprocating sliders that generate self-propelling motion.

Fig. 2: Bio-inspiration and conceptual locomotion principle of OSCAR. (a) Schematic representation of a parasitic wasp transporting eggs through coordinated stick–slip interactions generated by three sliding valves in their ovipositor. During each cycle, one valve retracts while the other two remain stationary, holding the egg in place because of the higher friction. When all valves reach their retract position, they advance together, producing a net forward displacement of the egg. Repeating this sequence enables controlled transport of the egg through the lumen till its deposition. (b) OSCAR translates and extends this principle into a capsule architecture using circumferential sliders driven by a rotating cam profile. Unlike the biological mechanism, where one valve moves at a time while the rest remain stationary, OSCAR drives a slider forward while driving the remaining sliders in the backward during each stroke. This continuous, phase-shifted reciprocation can yield a steady and continuous steadier average forward thrust suitable for self-propelling locomotion in colonoscopy.

Fig. 3: Design and prototype of the OSCAR capsule robot. (a) The design of OSCAR consists of 12 spring-loaded sliders distributed evenly around the circumference of the capsule body and driven by a central cam mechanism to realize the bio-inspired motion cycle described in the working principle. Each slider is guided by a cylindrical rod to ensure purely axial motion and is covered by a grooved slider cap to enhance frictional contact with the mucosa. The cam is mounted concentrically on a motor and enclosed by a back cover. Different cam profiles with varying numbers of motion reversals (jumps) were designed and fabricated to study the influence of motion pattern on thrust. (b) The spring-loaded cam system enables the use of steep cam angles while maintaining continuous contact between the sliders and the cam profile. This configuration allows maximum number of sliders to remain engaged in retraction, maximizing the potential net thrust. During the advance (fall) phase, the spring-loaded force of the advancing slider acting on the cam profile recycles the stored spring energy by generating a torque that assists the motor, thereby reducing the required actuation load. (c) Assembled and disassembled prototypes of OSCAR, showing all key components and the compact integration of the actuation and transmission mechanisms.

Design parameter	description	Value
$D_{capsule}$	Capsule diameter	$28~\mathrm{mm}$
$L_{capsule}$	Capsule length	$60~\mathrm{mm}$
$h_{stroke}$	Slider stroke	$10~\mathrm{mm}$
$m$	Mass	$50~\mathrm{g}$

TABLE I: Design parameters of the OSCAR prototype.

Fig. 4: Single–slider kinematics and thrust force generation. (a) Kinematic description of OSCAR’s slider actuation. A cam rotating with angular velocity

\omega

drives each slider through a prescribed linear trajectory

x(t)

. The cam law for the single–jump profile consists of a slow rise phase (slider retraction), a steep fall phase (slider advance), and a short dwell. The interval asymmetry between rise and fall sets the duty–cycle imbalance that underpins net thrust generation. (b) Thrust force model based on frictional interaction with the colonic wall. During retraction (top), the slider moves backward relative to the capsule, generating positive thrust as the wall traction opposes the slider motion. During advance (bottom), the slider moves forward, producing negative thrust as the wall reaction reverses direction. Superposition of these phase–shifted contributions of all sliders yields the average net thrust force.

Fig. 5: Simulation of kinematics and thrust force of OSCAR. (a) Single-slider displacement, velocity, and instantaneous traction force for 1-, 2-, and 3-jump cam geometries. Each cam produces distinct rise–return sequences corresponding to retracting (green) and advancing (red) phases, with normalized slider velocity given by $\dot{x}_{slider}^{*}/(\omega h_{\max})=\pm 1/(2\pi d_{r,a})$ . (b) Instantaneous and cycle-averaged capsule thrust force obtained by superposing the phase-shifted contributions of $n=12$ sliders. The upper plots show the total normalized thrust $\tilde{F}_{\mathrm{capsule}}(\theta)$ and its mean value $\bar{\tilde{F}}_{\mathrm{capsule}}$ , while the lower plots show the instantaneous number of retracting $n_{ret}(\theta)$ and advancing $n_{adv}(\theta)$ sliders. Increasing the number of jumps compresses the individual stroke, increases switching frequency, and reduces the average thrust, while amplifying ripple due to increased overlapping between slider transitions. Refer to Supplementary Video.1 for a demonstration of the motion cycles of the three cam profiles.

Fig. 6: Simulation of cycle-averaged thrust and instantaneous thrust ripple of OSCAR. (a) Duty-fractions map of the cycle-averaged normalized thrust $\bar{\tilde{F}}_{\mathrm{capsule}}$ over the cam duty fractions $(d_{ret},d_{adv})$ . The map shows that the mean thrust is set by the imbalance between the retract and advance intervals: $d_{ret}>d_{adv}$ yields forward (positive) thrust, $d_{ret}<d_{adv}$ yields backward (negative) thrust, and $d_{ret}=d_{adv}$ yields zero mean thrust. Markers indicate the three implemented cam profiles. The plot on the right shows that the normalized cycle-averaged thrust increases linearly with the duty asymmetry $\Delta d=d_{ret}-d_{adv}$ , independent of number of sliders $n$ (inset) and the number of cam motion reversals (jump count) $k$ . (b) Ripple of the normalized instantaneous thrust $\tilde{F}{\mathrm{capsule}}(\theta)$ , quantified as the coefficient of variation $\sigma{\tilde{F}{\mathrm{capsule}}}/\bar{\tilde{F}}{\mathrm{capsule}}$ , for different slider counts $n$ and jump counts $k$ . Ripple arises from overlapping slider phase switches that temporarily change the numbers of sliders in retraction and advance. The heat maps and representative curves show that $\Delta d$ shifts the phase boundaries within each of the $k$ segments relative to the slider spacing $2\pi/n$ , producing alternating low- and high-ripple regions as overlap is avoided or occurs. When $n/k$ is an integer (e.g., $n=12$ ), the transition pattern repeats across segments and ripple depends strongly on $(\Delta d,k)$ ; when $n/k$ is not an integer (e.g., $n=5$ ), the transition pattern becomes asymmetric and thus the ripple becomes less sensitive to $k$ .

Fig. 7: Experimental setups for characterizing the thrust force generated by OSCAR. (a) Single-slider force characterization. An ex-vivo porcine colon sample is opened and mounted flat on a silicone-backed bracket, which is guided horizontally and instrumented with a single-axis load cell to measure the traction force from the tissue side. The robot is mounted on a vertical guide so that a single slider presses against the tissue under controlled normal loading. This configuration enables direct observation of the stick–slip traction cycle and captures the viscoelastic response of the wall, which is essential for understanding the slider–tissue contact mechanics. (b) Full-robot thrust characterization. OSCAR is inserted coaxially into an intact porcine colon segment attached to a single-axis load cell. Measuring traction from the tissue side again allows tacking of the traction force on the wall which includes its viscoelastic response, linking individual-slider mechanics to whole-robot propulsion. This setup is further used to evaluate how actuation speed, cam duty cycle, and the number of motion reversals per cycle (cam jumps) influence the net thrust generated by the capsule.

Fig. 8: Single–slider friction experiment and validation of the viscoelastic stick–slip model. (a) Experimental setup and measured wall–force response during alternating retraction and advance of a single slider. The recorded traction force exhibits symmetric peak traction (

\pm 0.14

N) and clear stick–slip transitions shaped by the wall’s viscoelastic response. The zoomed-in view of the traction force profiles show the linear build-up during stick and the exponential convergence toward the friction limit during slip, together with the model prediction obtained using the experimentally-identified elongation stiffness

k_{tissue}\approx 130\mathrm{N.m^{-1}}

and viscous damping

c_{tissue}\approx 6\ \mathrm{N.s.m^{-1}}

parameters. The bottom panel shows error profile over two periods, with the model achieving

\mathrm{RMSE}=0.0101

N and

\mathrm{NRMSE}\approx 3.6\%

. (b) Estimated wall and slider kinematics and the corresponding stick–slip force evolution. The model reconstruction shows how slider and wall motion coincide during stick, producing linear traction growth, and how the slider force saturates during slip while the wall relaxes according to the tissue’s viscoelastic dynamics. These results demonstrate that the stick-slip formulation accurately reproduce the recorded interaction and reveal the underlying contact mechanics. Refer to Supplementary video.2 for a video demonstration of the single slider force characterization experiment.

Fig. 9: Capsule thrust force and dynamic imbalance. (a) Comparison between the measured traction force generated by the capsule on the ex-vivo colon wall (blue), the wall and capsule thrust force estimated by the viscoelastic model (grey and black), and the capsule thrust force estimated by the ideal model (the friction-limited case) (light green). The ideal model predicted cycle-averaged capsule thrust force ( $1.38~\mathrm{N}$ ) compared to ( $1.13~\mathrm{N}$ ) predicted by the viscoelastic model, which reflects the losses due to wall viscoelasticity. The absolute value of the experimentally measured steady-state average traction ( $\approx 0.85~\mathrm{N}$ ) is less than the absolute value of the wall force predicted by the viscoelastic model ( $1.05~\mathrm{N}$ ), attributed to the mechanical losses in the system. Insets show model–experiment comparison over one and two cam cycles, highlighting viscoelastic losses (top) and system-level mechanical losses (bottom). (b) Illustration of the inertial moment imbalance generated by sliders during the rapid advance phase. In the single-jump cam, only one slider undergoes the high acceleration motion reversal when switching to advance phase without any balancing counter force, producing a net rocking moment on the capsule and a large-amplitude ripple in the measured thrust. For higher jump counts, sliders advance in symmetric groups (two opposite sliders in the two-jump cam, three evenly spaced sliders in the three-jump cam), balancing the inertial moments and substantially reducing the thrust ripple. Refer to supplementary video.3 for a video demonstration of the full capsule force characterization experiment.

Fig. 10: Effect of speed and cam jumps on capsule thrust, and comparison between models and experiments. (a) Cycle-averaged traction forces generated by OSCAR under different voltages ( $6\mathrm{V},9\mathrm{V},12\mathrm{V}$ ) corresponding to different motor speeds using the single-jump cam, and under the three different cam geometries with ( $1,2,3$ ) cam jumps at $12\mathrm{V}$ driving voltage. Box plots and the corresponding time-evolution traces show the mean thrust remained approximately constant across the different speed, whereas increasing the number of cam jumps reduces the net traction and produces progressively smoother force profiles due to improved moment balancing across sliders. (b) Comparison between measured traction, the ideal friction model, the viscoelastic model, and the viscoelastic model with subtracted mechanical losses, plotted against duty-cycle asymmetry $\Delta d$ . Viscoelasticity lowers the ideal thrust, and additional mechanical losses observed in the full system shifts the prediction further downward, closely matching the experimental data. The lower-left panel shows corresponding ripple amplitudes. Right: the corresponding time-evolution traces for each cam design demonstrate how model and experiment agree in mean thrust while differing in ripple characteristics due to system-level inertial effects.

Fig. 11: ex-vivo validation of locomotion and tissue safety in porcine intestine. (a) Proof-of-concept experimental setup showing OSCAR alongside a long, excised porcine colon sample in its natural, collapsed configuration (without insufflation). (b) Time-lapse visualizing the progression of the capsule through the colon sample over a 55-second interval. The robot successfully propelled forward through the colon samples, achieving an average speed of $3.08~\mathrm{mm/s}$ . This speed corresponds to a theoretical full-length colon traversal time of $\approx 8$ minutes, comparable to the standard for conventional colonoscopy. Following locomotion, the intestinal tract was dissected (schematic inset) to assess the integrity of the tissue wall. (A–C) Representative microscopic views of the internal mucosal wall at distinct locations along the traversal path. The images reveal an intact epithelial surface with no visible abrasions, hematomas, or disruption, confirming the atraumatic nature of the ovipositor-inspired self-propelling mechanism. Refer to supplementary video.4 for video demonstration of the locomotion validation experiment.

Robotic Endoscopy

See all 17 papers

Jul 11, 2026

Soft Eversion Robots for Colonoscopy: Challenges, Open Problems, and Emerging Solutions

Conventional colonoscopy remains limited by patient discomfort and procedural risks, motivating research into compliant robotic alternatives. Eversion robots, which advance via pressure-driven tip growth, eliminate sliding friction against the colon wall and offer a less intrusive approach to traversal. However, no existing design simultaneously satisfies all clinical requirements. This paper benchmarks four recent eversion robot architectures against the key anatomical and clinical constraints of colonoscopy, including colon length, minimum luminal diameter, bending angle, and working-channel needs. We identify the central trade-offs each design reveals, particularly the difficulty of integrating steering and payload delivery without sacrificing the soft, low-friction behaviour that makes eversion robots attractive. We then provide design guidance across material selection, steering strategy, and payload delivery, and highlight open problems for clinical translation.

Cem Suulker, Thomas Mack, Qingzheng Cong +4

School of Engineering and Materials Science, Queen Mary University of London, United Kingdom.

Soft Robotics Surgical Robotics

Ask PDF

Apr 4, 2026

OpenRC: An Open-Source Robotic Colonoscopy Framework for Multimodal Data Acquisition and Autonomy Research

Colorectal cancer screening critically depends on colonoscopy, yet existing platforms offer limited support for systematically studying the coupled dynamics of operator control, instrument motion, and visual feedback. This gap restricts reproducible closed-loop research in robotic colonoscopy, medical imaging, and emerging vision-language-action (VLA) learning paradigms. To address this challenge, we present OpenRC, an open-source modular robotic colonoscopy framework that retrofits conventional scopes while preserving clinical workflow. The framework supports simultaneous recording of video, operator commands, actuation state, and distal tip pose. We experimentally validated motion consistency and quantified cross-modal latency across sensing streams. Using this platform, we collected a multimodal dataset comprising 1,894 teleoperated episodes ~19 hours across 10 structured task variations of routine navigation, failure events, and recovery behaviors. By unifying open hardware and an aligned multimodal dataset, OpenRC provides a reproducible foundation for research in multimodal robotic colonoscopy and surgical autonomy. Both the hardware design and dataset are available at https://github.com/artslab2019/openrc-robotic-colonoscopy. The dataset has been contributed to the Open-H-Embodiment Initiative by NVIDIA.

Siddhartha Kapuria, Mohammad Rafiee Javazm, Naruhiko Ikoma +4

Walker Department of Mechanical Engineering, The University of Texas at Austin, Austin, 78712, TX, USA · Department of Surgical Oncology, Division of Surgery, The University of Texas MD Anderson Cancer Center, Houston, 77030, TX, USA · Department of Surgery & Perioperative Care, Dell Medical School, The University of Texas at Austin, Austin, 78712, TX, USA +1

Robotic Endoscopy Surgical Robotics Robotic Data Collection Robot Teleoperation

Ask PDF

Sep 19, 2026

ARCGym: Benchmarking Deep Reinforcement Learning in Autonomous Robotic Colonoscopy

Simulations for learning-based autonomous colonoscopic navigation focus mainly on fully actuated capsule robots, failing to capture the contact-rich navigation of long and flexible clinical colonoscopes. We present the Autonomous Robotic Colonoscopy Gym (ARCGym), an open-source reinforcement learning environment and benchmark for image-based navigation in clinically derived deformable colon anatomies. ARCGym supports multiple types of colonoscope robots, spanning capsule robots and flexible endoscopes, with this work focusing on flexible endoscopes including magnetic-driven tip actuation and clinically used proximally translational actuation. This work includes five CT-reconstructed colons representing typical clinical scenarios, a set of clinically meaningful navigation subtasks, and unified success metrics. We introduce a reward combining depth-based lumen alignment with a lumen-visibility score to improve learning under occlusions. Experiments across tasks, robots, and anatomies show that autonomous navigation remains challenging for both magnetic-driven and proximal-insertion flexible robots, with proximal-insertion actuation remaining an open problem.

Guanglin Ji, Martina Finocchiaro, Kenny Erleben +1

Department of Computer Science, University of Copenhagen, Copenhagen, Denmark

Robotic Endoscopy RL Benchmarks Endoscopy Robotics Simulation Robotic RL Surgical Robotics

Ask PDF

Abstract

Robotic Endoscopy

Explore similar work

Soft Eversion Robots for Colonoscopy: Challenges, Open Problems, and Emerging Solutions

OpenRC: An Open-Source Robotic Colonoscopy Framework for Multimodal Data Acquisition and Autonomy Research

ARCGym: Benchmarking Deep Reinforcement Learning in Autonomous Robotic Colonoscopy