List of Posters

Snow removal is a critical task in cold regions, which entails challenges such as high labor intensity, safety risks, and inefficiency of traditional methods, particularly in aging populations and extreme weather conditions. The use of autonomous snow removal robots equipped with advanced navigation and snow recognition technologies has become an important approach to remove snow.

Li et al. proposed a novel adversarial reservoir computing (RC) method that enables safe navigation in snowy environments. While this method effectively avoids obstacles, it cannot identify snow-covered regions1. To address this issue, Kubo et al. proposed a deep learning-based semantic segmentation approach that combines LiDAR scan data and semantic maps. This method enables accurate recognition of snow regions, roads, and buildings2. However, it requires complex training data such as the labor-intensive annotation of semantic maps and has limited accuracy in identifying moving objects.

In our study, we proposed a navigation system that ensures safe navigation without obstacle collisions while guiding the robot toward snow-covered regions, thereby improving snow removal efficiency. Our system consists of two main components, navigation and segmentation. The navigation component uses thermal and distance images as input and employs a RC-based reinforcement learning optimized by artificial bee colony algorithm to generate the robot's velocity for obstacle avoidance. The segmentation component takes thermal and gray images as input and employs the k-means method to segment snow-covered regions, generating velocities with sigmoid functions to guide the robot.

We built an O-shaped track for the simulated snowy environment using Ignition Gazebo and used a four-wheeled robot to do experiments. The results demonstrate that our navigation method achieved continuous collision-free navigation in new environments, where the positions of obstacles were changed. The segmentation method successfully identified three types of regions including snow-covered regions, roads, and others, and achieved a mean accuracy of 95 %. The overall system successfully performed collision-free navigation with a snow removal coverage of 30 % just in one lap.

1 Li, Fangzheng and Ji, Yonghoon. "Dual-Type Discriminator Adversarial Reservoir Computing for Robust Autonomous Navigation in a Snowy Environment." IEEE International Conference on Ubiquitous Robots (2024): 713-717.

2 Kubo, Kohei and Emaru, Takanori. "Proposal of Voxel Semantics Estimation System for Automated Snowplow Operation." Proceedings of 2024 IEEE/SICE International Symposium on System Integration (2024): 502-507.

With the rapid increase in the utilization of service robots, enhancing their mobility and operational capabilities in diverse environments has become a crucial research area1. In particular, elevator button operation is essential for enabling service robots to autonomously move between floors, significantly expanding their practical applications. This study presents the design and implementation of a 4-degree-of-freedom (DOF) manipulator specifically developed for elevator button operation. A parallel mechanism2 was employed to eliminate the pitch DOF of the end-effector, thereby simplifying the structural design, reducing control complexity, and resolving singularity issues. Additionally, YOLO-based object detection technology was utilized to accurately extract the button's position from camera images, enhancing the precision of button operation. Future work aims to improve the manipulator’s autonomous movement by integrating an automatic correction algorithm based on pixel coordinate transformation3.

The proposed manipulator adopts a modular design, ensuring seamless compatibility with existing mobile robots and enabling easy deployment in various environments. To verify the reliability and practicality of the system, Multi-Body Dynamics (MBD) and Finite Element Analysis (FEA) were conducted to assess its structural stability and operating torque. Furthermore, forward and inverse kinematics analysis was applied to develop motion algorithms for precise button operation, ensuring accurate position control.

This study achieves both structural simplicity and cost efficiency compared to conventional elevator button operation robots while maintaining high reliability and operational precision. Furthermore, by integrating AI-based object recognition and autonomous control technologies, this research suggests the potential for a fully automated elevator operation system. The proposed manipulator is expected to contribute to expanding the application range of service robots and serve as a foundation for broader use in future automation systems.

1 GONZALEZ-AGUIRRE, Juan Angel, et al. Service robots: Trends and technology. Applied Sciences, 2021, 11.22: 10702.

2 BARKER, Clark R. A complete classification of planar four-bar linkages. Mechanism and Machine Theory, 1985, 20.6: 535-554.

3 MARDIATI, Rina, et al. The derivation of matrix transformation from pixel coordinates to real-world coordinates for vehicle trajectory tracking. In: 2019 IEEE 5th International Conference on Wireless and Telematics (ICWT). IEEE, 2019. p. 1-5.

With the development of the space industry, there is an increasing demand for deployable space structures with high efficiency and reliability, such as spaceborne antennas for satellite communication, high-resolution Earth imaging, and deep space exploration1. Although conventional parabolic and ring truss antennas have been widely used, they still face limitations in deployment reliability and stowage efficiency2. To address these issues, the primary objective of this work is to minimize mechanical complexity while providing a synchronized deployment by utilizing a pantograph rib structure. Specifically, we present a novel deployable spaceborne mesh antenna mechanism integrating a pantograph rib structure. The proposed antenna consists of eight pantograph ribs that are deployed using a single actuator. Unlike conventional designs, the mesh reflector is directly attached to the ribs, which does not require additional cable networks. To maintain axial stiffness and ensure synchronized deployment, SMA (shape memory alloy) wires are employed. The antenna structure deploys through a linear screw-driven motor. To validate the deployment of the proposed structure, a prototype was fabricated. In parallel, multi-body dynamics simulations were conducted by utilizing RecurDyn. A modal analysis using MSC Patran/Nastran was performed to assess the impact of SMA wires on structural stiffness. Additionally, to verify deployment reliability, ten repeated deployment tests were conducted following ECSS standards3. As a result, the aperture diameter of the antenna mechanism increases with a deployment ratio of 7:1, and a volume deployment ratio of 58.8:1. These results represent that the proposed antenna ensures a high stowage efficiency compared to conventional antenna designs. Moreover, the experimental results identified synchronized deployment for all components. Modal analysis revealed that the natural frequency increased by 1.42 times when SMA wires were applied, resulting in approximately double increase in structural stiffness. In a nutshell, this study presents a new deployable antenna mechanism that balances high stowage efficiency with deployment reliability. It must be highlighted that the limitations of existing antenna designs have been successfully addressed by combining the pantograph rib structure with SMA wire reinforcement.

1 Lappas, V., & Kostopoulos, V. "A survey on small satellite technologies and space missions for geodetic applications." Satell. Mission. Technol. Geosci. 8 (2020): 123-144.

2 Duan, Baoyan, Yiqun Zhang, and Jingli Du. Large Deployable Satellite Antennas. Springer, 2020.

3 ECSS-E-ST-33-01C Rev.1 – Mechanisms (15 February 2017) European Cooperation for Space Standardization (ECSS), 2017.

In recent trends in the automotive industry, the electrification of various components has been widely adopted to enhance user convenience. However, in high-end vehicles, even minor noises from these systems can negatively affect the perceived quality.1 Durability testing is typically conducted to ensure reliability against such issues prior to market release. Squeak noises from power door systems are a well-known example. However, the inspection is usually based on on-site hearing by humans in noisy and uncontrolled environments, making the process both unreliable and subjective. This paper presents a novel method for diagnosing squeak noise using accelerometer signals. In contrast to microphone-based approaches, the proposed diagnostic model minimizes external noise interference. To improve diagnostic performance, a psychoacoustic feature parameter3, which quantifies the subjective perception of sound, was applied to the accelerometer signal, demonstrating the method’s effectiveness across various squeak types. To assess the constructed model, both accelerometer and acoustic signals were collected under durability test conditions, and their diagnostic performance was compared. The results indicate that the proposed methodology demonstrates similar or superior diagnostic performance compared to using acoustic signals, suggesting that this approach can be effectively extended to diverse operational environments where squeak detection is essential.

1 Nolan, Stephen A., and Joseph P. Sammut. Automotive squeak and rattle prevention. No. 921065. SAE Technical Paper, 1992.

2 Zio, Enrico. "Prognostics and Health Management (PHM): Where are we and where do we (need to) go in theory and practice." Reliability Engineering & System Safety 218 (2022): 108119.

3 Zwicker, Eberhard, and Hugo Fastl. Psychoacoustics: Facts and models. Vol. 22. Springer Science & Business Media, 2013.

The dexterous manipulation capabilities of humans are attributed to the hybrid soft-rigid structures of the fingers and tactile perception. Recent advancements in Vision-Based Tactile Sensors (VBTS), particularly through data-driven methods, have enabled robotic hands to incorporate tactile perception. However, most VBTS research has focused primarily on perception, without replicating the structural complexity of the human hand. In this paper, we introduce a fingernail-inspired structure, which mimics the nail and bone of human fingers, into a soft-material based VBTS. Our results show that this addition not only enhances the contact area but also increases the grasping force with objects compared to designs lacking fingernail features. 

This study is intended to develop a simulation environment for colonoscopy procedures utilizing SOFA and Isaac Sim. The aim of this study is to develop a realistic colonoscopy simulation model using FEM and SOFA to enhance the accuracy of endoscopic manipulation, enable precise collision detection, and ultimately improve the effectiveness of medical training. The objective is to accurately replicate the complex physical interactions that occur during actual colonoscopy procedures in a virtual environment, thereby enhancing the precision of endoscopic manipulation and facilitating the development of an effective simulator for operator training and education.

Recent research trends in medical simulation have focused on improving the accuracy of physics-based deformable body simulations. In particular, implementing realistic simulations through real-time feedback and high-performance physics engines has emerged as a critical challenge. These technological advancements significantly contribute to maximizing the effectiveness of medical education and training1.

In terms of methodology, an endoscope-like model was developed, and both the endoscope and colon models were constructed using the finite element method (FEM) based on the SOFA framework. The endoscope which I made was modeled as the front section of a beam structure, with movement and rotation controlled via joystick input through ROS2 communication. Collision detection was implemented using geometric collision models, while the colon model was configured as a deformable body, incorporating both self-collision and external collision detection functionalities.

Principal experimental results demonstrated that real-time control of the endoscope via joystick was physically realistic, with stable data transmission verified through ROS2 bridge. Various beam structures were successfully controlled in real-time, confirming the feasibility of endoscopic navigation within the colon. However, maintaining real-time performance due to computational overload in the SOFA physics engine is a challenge.

For future work, I plan to address the physics engine overload issues by leveraging Isaac Sim to create a more realistic and sophisticated simulation environment. Additionally, I will explore the application of reinforcement learning and imitation learning techniques to optimize performance and develop effective control strategies within complex colon environments. Ultimately, my goal is to apply these research outcomes to real-world scenarios through Sim2Real studies, advancing actual colonoscopy navigation technologies.

1 Elendu, Chukwuka, et al. "The impact of simulation-based training in medical education: A review." Medicine 103.27 (2024): e38813.

Soft robots, compared to traditional rigid robots, offer inherently safe and adaptive interaction with delicate objects and humans in dynamic environments due to their deformability, compliant attributes1. Various approaches have been introduced to develop soft actuators. (i.e., fluids, cables, electric, polymers etc.). Fluidic actuators can exhibit omnidirectional deformation depending on their structural design upon the imposed pressure2. Among them, pneumatic soft actuators stand out due to their light weight, safety, low cost and ease of fabrication3. With these advantages in mind, we propose a novel pneumatic driven fingerless soft gripper with pleat patterns. Unlike prosthetic fingered gripper our fingerless gripper can engage irregular objects without the prior knowledge on the object, decision-making, or a specific control strategy in unconstructed environments4. In our approach, morphological and geometric studies were carried out in order to compromise deformability and flexibility. To achieve desired kinematics, patterns are strategically implemented for omnidirectional deformation by the structure rather than relying on the properties of the material. The FEM results showed that the pleat patterns cause greater deformation compared to flat cylinders, while the outer walls thicker than inner walls exhibit greater holding strength. Incorporating the fingerless soft gripper with a commerical 6 DoFs manipulator, we demonstrated various scenarios, showing strong potentials in real-world applications which includes pick and place of arbitrary shape objects. In light of these demonstrations, it is worth to noting that the fingerless soft gripper can be used for food robotics, sorting and aligning system via multi-dimensional motions. To develop further towards soft machines that can interact with unconstructed environment, our future work will be to integrate proprioceptive sensors and to realize sensory feedback.

1 Tawk, Charbel, and Gursel Alici. "A review of 3D‐printable soft pneumatic actuators and sensors: research challenges and opportunities." Advanced Intelligent Systems 3.6 (2021): 2000223.

2 Joe, Seonggun, Federico Bernabei, and Lucia Beccai. "Chapter A Review on Vacuum-Powered Fluidic Actuators in Soft Robotics." (2022).

3 Gu, Guoying, et al. "Analytical modeling and design of generalized pneu-net soft actuators with three-dimensional deformations." Soft robotics 8.4 (2021): 462-477.

4 Jain, Snehal, et al. "A multimodal, reconfigurable workspace soft gripper for advanced grasping tasks." Soft Robotics 10.3 (2023): 527-544.

Interest in actively controlling droplet movement has been increasing, and research utilizing electrowetting, magnetic fields, and vacuum pressure to manipulate droplets is actively being conducted. However, these methods require additional external energy input, which can lead to challenges such as complex system configurations, high costs, reduced energy efficiency, and limited applications in soft materials. In contrast, we propose a three-dimensional patterned design that enables directional flow without external energy1. This structure was fabricated by ultra-precision machining an aluminum (Al) substrate to create the three-dimensional pattens, followed by dip-coating. Under specific conditions, superhydrophobic characteristics were observed, confirming that droplets could exhibit directional movement on the superhydrophobic pattern without external energy.

1 Wang, Shun, et al. "Directed Motion of an Impinging Water Droplet—Seesaw Effect." Journal of Materials Chemistry A 8.13 (2020): 6360-6370.

Satellite antennas play a crucial role in various space missions, including mobile communications, GPS, Earth observation, and space exploration1-3. In particular, with the rapid advancement of the space industry in recent years, the importance of high-resolution satellite antennas has been increasing. To achieve precise observations and efficient data transmission, antennas with large apertures are essential4,5. However, such large satellite antennas not only increase the satellite’s weight at launch but also add to the burden on the launch vehicle and escalate launch costs. Despite these challenges, efficiently stowing these large antennas within the constrained payload space and ensuring their stable deployment during missions remain persistent engineering challenges6,7. Moreover, the design of deployment mechanisms for large antennas must ensure lightweight characteristics, a high deployment ratio, and post-deployment stability4,8,9.

In this study, a one-degree-of-freedom (1-DOF) pantograph structure is adopted and applied to a 6R ring truss antenna to achieve synchronized motion. Additionally, structural analysis is conducted on the improved deployment mechanism to examine stress distribution and deformation, and its feasibility is verified through comparison with a conventional 6R ring truss antenna. To evaluate the mechanical performance of the deployment mechanism in dynamic environments, multi-body dynamics analysis is performed to investigate stress distribution and deployment feasibility.

1 Kopacz, Joseph R., Roman Herschitz, and Jason Roney. "Small satellites an overview and assessment." Acta Astronautica 170 (2020): 93-105.

2 Kodheli, Oltjon, et al. "Satellite communications in the new space era: A survey and future challenges." IEEE Communications Surveys & Tutorials 23.1 (2020): 70-109.

3 Zhang, Yiqun, et al. "Dynamic analysis of the deployment for mesh reflector deployable antennas with the cable-net structure." Acta astronautica 131 (2017): 182-189.

4 Duan, Baoyan. "Large spaceborne deployable antennas (LSDAs)—a comprehensive summary." Chinese Journal of Electronics 29.1 (2020): 1-15.

5 Rahmat-Samii, Yahya, and Arthur C. Densmore. "Technology trends and challenges of antennas for satellite communication systems." IEEE Transactions on Antennas and Propagation 63.4 (2014): 1191-1204.

6 Kang, Hyeongseok, Bohyun Hwang, Sooyoung Kim, Hyeonseok Lee, Kyungrae Koo, Seonggun Joe, and Byunkyu Kim. "A Conceptual Design of Deployable Antenna Mechanisms." Aerospace 11.11 (2024): 938.

7 S. Kim, H. Kang, R. H. Do, K. Koo, and B. Kim, "Design of large deployable mechanism with frustum shape satellite antenna," Journal of Aerospace System Engineering, vol. 18, no. 6, pp. 64-70, 2024, doi: 10.20910/JASE.2024.18.6.64

8 Wang, Bing, et al. "Space deployable mechanics: A review of structures and smart driving." Materials & Design 237 (2024): 112557.

9 Qi, Xiaozhi, et al. "A large ring deployable mechanism for space satellite antenna." Aerospace Science and Technology 58 (2016): 498-510.

As autonomous driving gains widespread attention, extensive research is being conducted to enable robots to safely reach their destinations independently. Before the rise of advanced robotic artificial intelligence, much research focused on path planning for route exploration and planning. However, these methods have limitations in responding to dynamic environmental changes. To overcome this, efforts have been made to integrate reinforcement learning. This study proposes an adaptive path planning reinforcement learning algorithm, combining the efficiency of path planning with the adaptability of reinforcement learning to various environmental changes. The proposed method effectively integrates traditional path planning algorithms with reinforcement learning. We utilize the TD3 network as the backbone and apply the Artificial Potential Field algorithm, known for its robustness to obstacles. Experimental results demonstrate that our approach achieved approximately 14 points higher average rewards and similar maximum rewards compared to existing networks, indicating higher average success rates. Additionally, resulting in faster or comparable convergence speeds relative to traditional networks.

1 J Jung, C. Y., Kim, T., & Shim, D. H. (2019). Development of ros-based mobile robot system and experiments on indoor autonomous driving. Journal of Institute of Control, Robotics and Systems, 25(5), 438-444.

2 Fujimoto, S., Hoof, H., & Meger, D. (2018, July). Addressing function approximation error in actor-critic methods. In International conference on machine learning (pp. 1587-1596). PMLR

3 Park, M. G., Jeon, J. H., & Lee, M. C. (2001, June). Obstacle avoidance for mobile robots using artificial potential field approach with simulated annealing. In ISIE 2001. 2001 IEEE International Symposium on Industrial Electronics Proceedings (Cat. No. 01TH8570) (Vol. 3, pp. 1530-1535). IEEE.

4 Souissi, O., Benatitallah, R., Duvivier, D., Artiba, A., Belanger, N., & Feyzeau, P. (2013, October). Path planning: A 2013 survey. In Proceedings of 2013 international conference on industrial engineering and systems management (IESM) (pp. 1-8). IEEE.

With the widespread adoption of home 3D printers, the digital fabrication technique of turning 3D models into real objects has become more accessible. However, creating robotic arms remains challenging for common users. The simulation of a robotic arm’s movements requires kinematic knowledge, and mastering 3D modeling software can be time-consuming, both of which necessitate specialized expertise and skills.

To address these issues, this study developed a robotic joint mechanism design system, Sketch2Joints, which can automate the design, simulation, and 3D model creation processes through a sketch-based interface. With the help of the proposed system, users can design robotic arms by performing simple operations, such as clicking and dragging on a canvas within the user interface. The movement of the designed robotic arm is then simulated using the Rodrigue’s rotation formula for arbitrary axes. Additionally, the proposed system can save the designed robotic arm as a 3D model for outputs that can be fabricated using a common desktop 3D printer, allowing users to obtain a 3D model of robotic joints without the use of commercial 3D modeling software. This proposed system enables the design and fabrication of robotic arms without requiring specialized knowledge and skills. The utility of this system was demonstrated through user experiments conducted with non-expert users.

Autonomous snow-removal robots require precise environmental perception to operate effectively in dynamic winter conditions. This study presents an integrated framework that combines drone-based data collection, geostatistical interpolation, and simulation to improve snow depth mapping and robotic navigation. A UAV-mounted system integrates RGB imaging and LiDAR scanning to capture high-resolution spatial data over snow-covered areas. These datasets are fused to generate a 3D mesh model with high geometric accuracy, specifically designed for snow environments and comparable to established methods. For snow depth estimation, LiDAR point clouds provide structural depth information, while RGB images assist in detecting snow boundaries, enhancing accuracy beyond single-sensor approaches.

The system builds on SLAM-based 3D mapping1 and employs the geospatial sampling approach, Kriging interpolation2 to improve sparse data sampling efficiency. The reconstructed 3D mesh serves as a simulated environment for testing robotic navigation under various snow conditions. By linking snow depth variations to robotic mobility constraints, the proposed path planner optimizes snow redistribution, obstacle avoidance, and snow removal strategies.

Future work will focus on adapting this framework for real-world robotic deployment, with continued research into real-time sensor fusion under more complex environmental conditions. This study bridges high-precision environmental modeling with autonomous operation, contributing to advancements in winter maintenance technologies.

1 Lou, Lu, Li, Yitian, Zhang, Qi, and Wei, Hanbing. "SLAM and 3D semantic reconstruction based on the fusion of Lidar and monocular vision." Sensors 23.3 (2023): 1502.

2 Zimmermann, Niklaus E., Edwards, Thomas C., Jr., Graham, Catherine H., Pearman, Peter B., and Svenning, Jens-Christian. "New trends in species distribution modeling." Ecography 33.6 (2010): 985-989.

Due to the flexibility and material nature of soft robotics, soft robot arms are thought to be suitable for accelerating the fields of human augmentation and human-robot interaction without harming the human body. It is a promising way to provide a novel means of communication and body expression with the wearable robotic devices. However, soft robots usually require careful selection of materials and actuators, and intricated design of structure to solve these issues. This work aims to use biomimetics enabling a scalable, rapid and inexpensive fabrication process of such device. Therefore, we propose SpiMan, a wearable robotic device with logarithmic spiral-shaped tentacle-type structure, which is implemented based a soft manipulator inspired by SpiRobs1 that enables flexible and organic movements. In addition, a Rhinoceros-based design support system was developed to facilitate the fabrication of SpiMan. 

1 Z. Wang, N. M. Freris, and X. Wei. Spirobs: Logarithmic spiral-shaped robots for versatile grasping across scales. Device, p. 100646, 2024.

The proboscis trunk represents a multifunctional organ capable of versatile and dexterous grasping and manipulation. Among mammalian animals, it is remarkable that the proboscis trunk can exhibit an infinite degree of freedom movements while carrying hundreds of weights of the object1. Noteworthy, it must be noted that the elephant trunk has a specific set of motion primitives (i.e., elongation and contraction, torsion, bending, etc.), which plays a vital role in achieving muscular synergic motions2. Inspired by multifunctionality of the proboscis trunk, this work addresses pivotal aspects from biology to engineering and poses a bottom-top approach to mimic the true nature of the elephant. Specifically, the primary objectives of this work are to understand principles behind multifunctionality of the elephant trunk, and to propose a mechanical system ensuring multimodality. With these in mind, this work presents a bio-inspired pneumatic actuator in which is in the form of underactuated system. The bio-inspired pneumatic actuator (BIPA) employes a bellow skin (having convolutions) and aims to achieve multimodal grasping as well as high energy efficiency and to integrate intelligent systems via proprioception. The BIPA is composed of inlet and outlet – once the pressure imposed (either, positive or negative), the pressure is then being differentiated, resulting in gradient in pressure. Indeed, such gradient pressure is mainly due to the orifice effect, which has been widely used in differential nozzles3. In our approach, we performed CFD analysis using ANSYS and identified the gradient pressure when the inlet and outlet diameters were 3.8mm, and 0.819 mm, respectively. More specifically, it is identified that at the fixed flowrate of 360,000 mm3/min, as the difference in diameters of inlet and outlet becomes larger, the gradient pressure becomes significant. We then extended this work to fabricate prototypes via both molding and 3D printing technology and implemented adaptive-robust feedback control using PID-sliding mode controller (PID-SMC). As a result, the linear motion of the BIPA can be precisely controlled via PID-SMC. Particularly, it is worth noting that upon vacuum pressure both contraction and suction can be accomplished.

Given that many soft pneumatic actuators are vulnerable to cyclic operations due to poor resilience and reliability, it must be highlighted that presented approaches are highly useful in ensuring sustainable soft pneumatic actuators having material failures by employing PID-sliding mode control.

In light of these findings, our future work is to develop a specific pneumatic control setup that can accommodate high flowrate, resulting in large gradient in pressure and to achieve bio-inspired robots integrating muscular synergic motions and strategies via proprio- and exteroceptive sensory feedback together with adaptive control.

1 Dagenais, Paule, et al. "Elephants evolved strategies reducing the biomechanical complexity of their trunk." Current Biology 31.21 (2021): 4727-4737.

2 Joe, Seonggun, et al. "Jointless bioinspired soft robotics by harnessing micro and macroporosity." Advanced Science 10.23 (2023): 2302080.

3 Aguirre-Mendoza, Andres M., et al. "Effects of orifice sizes for uncontrolled filling processes in water pipelines." Water 14.6 (2022): 888.

Optical-based soft haptic sensing interfaces utilizing cameras offer a simple design with minimal wiring and maintenance requirements. Additionally, because the flexible outer skin of vision-based tactile sensors remains mechanically uncompromised, these sensors demonstrate high reliability and resistance to noise. However, as the sensing area expands, multiple cameras are typically required behind the soft skin to track marker displacements. Previously, we developed a barrel-shaped tactile sensor incorporating two cameras positioned at either end to extract three-dimensional (3D) marker movements. However, processing multiple camera images simultaneously demands significant computational resources. 

In this presentation, I will show a 3D estimation method that leverages a single fisheye camera rotating within the boundary of the soft skin, driven by a simple servo motor. By adjusting the camera’s rotation, sensing time and sensitivity can be actively controlled. This approach not only optimizes the usable volume within the sensor but also enhances measurement robustness against deformations of the outer skin1.

1 S. Nagasawa and V. A. Ho, "Three-Dimensional Shape Construction in a Soft Large-Scale Vision-based Tactile Sensor with a Single Rotational Camera," 2023 IEEE International Conference on Soft Robotics (RoboSoft), Singapore, Singapore, 2023, pp. 1-6, doi: 10.1109/RoboSoft55895.2023.10121933.