Advanced Robotics

Matsumoto, Ojiro; Tanaka, Hiroaki; Tadakuma, Kenjiro; Hosoda, Koh

doi: 10.1080/01691864.2025.2528829pmid: N/A

The body structure of living organisms provides a useful insight into sophisticated designs for robots. Incorporating a spine into quadruped robots has become a trend, with implementations of trunk flexion and torsion through the spine aimed at improving locomotion performance. However, the integration of multidirectional spine shape change requires a streamlined design to minimize mechanical complexity. This study introduces a spine structure inspired by the latissimus dorsi muscles that wrap around the trunk of quadruped animals, connecting the dorsal trunk to the forelimbs. The proposed design achieves two functions: generating trunk torsion through differential contraction of left and right latissimus dorsi muscle actuators, and converting spine flexion into forelimb propulsion motion via the latissimus dorsi muscle. We develop a forward kinematics model that incorporates a three-dimensional muscle path to estimate the design parameters achieving the intended functions in a real robot. Experiments with a quasi-quadruped robot based on the kinematic model show that the proposed design effectively integrates these functions using the latissimus dorsi muscle actuator.

Kambara, Motonari; Sugiura, Komei

doi: 10.1080/01691864.2025.2532610pmid: N/A

In this work, we address the problem of predicting the future success of open-vocabulary object manipulation tasks. Conventional approaches typically determine success or failure after the action has been carried out. However, they make it difficult to prevent potential hazards and rely on failures to trigger replanning, thereby reducing the efficiency of object manipulation sequences. To overcome these challenges, we propose a model, which predicts the alignment between a pre-manipulation egocentric image with the planned trajectory and a given natural language instruction. We introduce a Multi-Level Trajectory Fusion module, which employs a state-of-the-art deep state-space model and a transformer encoder in parallel to capture multi-level time-series self-correlation within the end effector trajectory. Our experimental results indicate that the proposed method outperformed existing methods, including foundation models.

Acs, Matthew; Zhong, Xiangnan

doi: 10.1080/01691864.2025.2532606pmid: N/A

This paper investigates the performance of behavior cloning (BC) with RGB inputs in the context of surgical robotics, focusing on data efficiency and generalization capabilities. Utilizing the LapGym ReachEnv simulation, we trained agents to perform a 3D reaching task using 2D visual data and conducted a dataset ablation study to determine the minimal data requirements for achieving optimal performance. The results show that BC achieves a 100% success rate with 25,000 episodes (∼1.23 million frames) of expert demonstrations, while performance significantly declines with smaller datasets. Additionally, we compare BC with offline reinforcement learning (RL) using Twin Delayed Deep Deterministic Policy Gradient with Behavior Cloning (TD3 + BC) and online RL using Proximal Policy Optimization (PPO), each evaluated in a sparse reward setting. TD3 + BC demonstrated superior generalization across more stringent task conditions, while PPO failed to learn a successful policy. Our findings suggest that offline RL, particularly TD3 + BC, offers improved robustness and generalization compared to supervised BC alone, especially in challenging surgical tasks. These insights highlight the potential of image-based learning and offline RL techniques in robot-assisted surgery, where sparse rewards and visual inputs dominate, and expert data is available.

Kayawake, Ryotaro; Abe, Kazuki; Tadakuma, Kenjiro

doi: 10.1080/01691864.2025.2523867pmid: N/A

In robotics, locking mechanisms are crucial in reducing energy consumption of devices such as manipulators. One type of locking mechanism that is widely utilized is the singularity-type locking mechanism. This device is designed such that its degree of freedom decreases at the singular point, allowing it to support large external forces for extended periods with minimal energy input. However, conventional singularity-type locking mechanisms face challenges in designing singular points. In this study, we propose a novel singularity-type locking mechanism based on rolling spool dynamics (called spool paradox). This innovative device applies a curved trajectory to the spool paradox, enabling the arbitrary design of singular points by adjusting the shape of the curve. Through a developed theoretical model and experiments, we demonstrated the capability of our proposed mechanism in locking external forces with minimal energy input. Furthermore, we applied the proposed mechanism to 1-DOF robotic arm and confirmed its motion.

Arakelian, Vigen

doi: 10.1080/01691864.2025.2528827pmid: N/A

Designing manipulation systems with predetermined initial and final positions of the gripper is an important field in modern robotics. The objective is to move the payload along a non-imposed trajectory between two given positions while allowing periodic position changes. This study proposes advanced design concepts for planar 5R parallel manipulators. The proposed design principle focuses on interconnecting the two links of the manipulator, ensuring the gripper’s initial and final positions. The mechanical system employs only one actuator, leading to simplified control and minimal energy expenditure. Consequently, the operational reliability is improved, and the overall cost is reduced. Two different design concepts are discussed: one involves the addition of gears, and the other employs the synthesis of a four-bar linkage with adjustable link lengths. Numerical simulations are conducted to demonstrate and validate the effectiveness of the proposed design concepts. This work presents a novel scientific contribution by introducing a new design concept: it explores the field of fixed-sequence manipulators built on the base of planar 5R parallel mechanisms. A design concept that has hitherto remained unexplored. Furthermore, the significance of this achievement is underscored by the attainment of an explicit mathematical solution, making it more attractive. This approach widens the scope of designing methods for fast manipulation systems, thereby expanding their practical applications in robotics.

Sakamoto, Kosuke; Koeba, Yutaro; Kunii, Yasuharu

doi: 10.1080/01691864.2025.2530515pmid: N/A

This paper presents a collision-based probabilistic Vector Field Histogram (p-VFH) obstacle avoidance algorithm for swarm robot navigation in unknown environments. Conventional obstacle avoidance strategies, including well-known path planning methods like A* and RRT*, are ill-suited for unknown environments. Moreover, current collision avoidance approaches for robot swarms face challenges related to computational demands, sensor performance, and potential local minima issues, such as deadlocks. Our proposed p-VFH algorithm tackles these problems by employing a probabilistic approach to determine the robot's movement trajectory. This method relies on a dynamically updated polar histogram that represents obstacle density in the surrounding area, and target direction. The algorithm begins by initializing histogram values, which are then updated as the robot encounters unrecorded obstacles. Subsequently, it creates a probability distribution to guide the selection of the next movement direction. To assess the effectiveness of p-VFH, we conducted comprehensive simulation studies. These experiments compared p-VFH's performance against three alternative methods including conventional VFH, a combined probability distribution approach, and a constant weighting function. The results show that p-VFH significantly improves exploration efficiency, successfully guiding robots to designated targets while effectively avoiding obstacles. In particular, p-VFH outperformed the other tested methods in terms of success rates and environmental adaptability. Furthermore, we conducted real-world experiments using a swarm robots equipped with the p-VFH algorithm. These real-world tests confirmed the effectiveness of p-VFH in real-time obstacle avoidance and exploration in unknown environments. The promising results suggest that the p-VFH algorithm could play a crucial role in advancing swarm robotics technology, with potential applications ranging from planetary exploration to various other fields.

Sato, Eiki; Maruta, Ichiro; Fujimoto, Kenji

doi: 10.1080/01691864.2025.2526598pmid: N/A

This paper presents a new approach to quadrotor state estimation and control using ultra-wideband (UWB)-based bidirectional ranging and angle of arrival (AoA) measurements. This method offers a significant advantage over traditional motion capture systems or time-of-arrival (ToA) based positioning by requiring only a single, compact UWB anchor, greatly simplifying setup and reducing infrastructure requirements. We address the challenge of significant noise in UWB measurements by integrating them with an inertial measurement unit (IMU) data through an extended Kalman filter (EKF). Our key contributions also include proposing a magnetometer-free yaw estimation method utilizing bidirectional AoA measurements. This method effectively addresses gyroscope drift in indoor environments where magnetometers are unreliable. We also provide theoretical validation of the system's observability, and experimentally demonstrate successful stabilization control of a micro quadrotor using the estimated states. The experimental results show significant reduction in estimation errors compared to raw sensor data. Additionally, we conduct control experiments in rainy environments and confirm that UWB-based control can be effective in a wide range of conditions, including rainy weather. Our approach offers a robust, cost-effective solution for quadrotor navigation and control in GPS-denied environments, particularly indoors, and in rainy conditions, while minimizing setup complexity and hardware requirements.