In this study, we investigated the effect of arm swinging with a flexible trunk on bipedal locomotion using a simulation model. The model imitates human arm swinging and flexible trunk twisting. The model showed that the center of pressure (CoP) oscillates sinusoidally in the left-right direction by asymmetrical arms swinging. We compared the stability of bipedal locomotion without arm swinging, with symmetrical arm swinging, and with asymmetrical arm swinging, by applying external force. The results showed that asymmetric arm swinging did not lead to fall over with a wider disturbance range and the CoP trajectory gradually became cyclic after the external force was applied.
Fisheye cameras' broad field of view makes them ideal for a variety of applications, such as robotic navigation, drone vision, and car surround-view systems. Fisheye eyeglasses' intrinsic nonlinear distortion, however, makes it difficult to accurately interpret images and analyze them later. This research proposes a Zynq System on Chip (SoC)-based accelerator to dynamically restore fisheye lens image distortion in real time. The proposed system improves the limitations of conventional field programmable gate array (FPGA) designs that require fixed parameters and leverages the Python Productivity for Zynq (PYNQ) framework to enable dynamic parameter modification without reprogramming. Experimental results show that the proposed system processes images with a resolution of 960x640 at a speed of 135.7 frames per second (fps), achieving more than 20 times faster performance compared to software-based solutions. Additionally, it can maintain real-time performance even at Full HD resolution. We also evaluated the accuracy by comparing the fisheye distortion correction of four interpolation techniques provided by OpenCV with the quality indicators Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). It is also flexible enough to allow the system to dynamically adapt to environmental changes and process images efficiently. This research provides a highly flexible, real-time fisheye distortion correction system for various industries such as robotics, automotive, and drones.
This paper presents the design and development of a novel, low-cost, fixed-base humanoid robot equipped with a torso, two arms, hands with five fingers, and a head. One of the biggest challenges in developing low-cost humanoid robots is identifying optimal ways to minimize manufacturing and production costs without compromising quality or performance. This requires creative engineering and design strategies and using cost-effective and readily available materials and components. The design prioritizes the use of fewer components, including a reduced number of sensors and actuators, to lower costs. Another approach involves manufacturing the robot's body parts using 3D printing technology, which reduces the costs associated with traditional manufacturing methods. As a result, it allows the robot to meet specific demands at any stage of the process economically. During the design process, careful consideration is given to selecting materials, kinematic structures, and mechanical components. Choosing suitable materials is crucial when designing a low-cost humanoid robot. Using inexpensive, strong, and lightweight materials is essential to keep the robot's cost down without compromising essential features. The robot's range of motion is dictated by its skeleton or kinematic structure. Subsequently, selecting mechanical parts, such as joints, actuators, and sensors, is critical in determining the robot's overall performance and agility. In the development process, building the physical body of a 6-DoF humanoid robot, integrating sensors for perception, and designing control and decision-making software are all necessary steps. When these components come together, they enable the creation of a robot capable of human-like movement and interaction. The ability of a humanoid robot to identify and interact with humans has significant real-world applications, especially in security roles where the robot can recognize individuals in public spaces. The humanoid robot is designed with 6 degrees of freedom (DoF) for its main body and an additional 5 DoF for each hand. In this design, the robot is equipped with an AI camera for facial recognition, enabling it to detect and identify human faces.
Investigated how bacteria self-propel in low Reynolds number liquid environments and modeled the swimming state. Through dimensional analysis, a centimeter-sized bionic robot, driven by a rubber band simulating a flagellar motor, was designed and fabricated. Kinematic experiments were conducted to test the swimming speed of the robot in liquids of different viscosities and to analyze the relationship between the swimming speed and the flagellar rotation speed. The experimental results show that the bio-inspired robot can mimic the continuous swimming motion of bacteria driven by elastic forces with a maximum speed of 25 mm/s in water. These results provide further insights into the study of micro bionic robots.
In contemporary manufacturing, efficient operation of robotic cells is critical and Job Shop Scheduling (JSS) faces challenges such as machine failures and shifting priorities. Deep Reinforcement Learning (DRL) addresses these challenges by enabling adaptive scheduling. DRL empowers robotic cells to dynamically optimize schedules which can increase resilience drastically. Through simulations and real-world use, DRL-based approaches outperform traditional methods, improving adaptability and efficiency. Integrating DRL into scheduling processes enables better navigation of uncertainties, enhancing productivity and competitiveness in modern manufacturing. This contribution focuses on these problems and provides a DRL approach for robotic cells to dynamically schedule production orders when facing random tool breakdowns. Using the epsilon-greedy policy, it allows the agent to explore the environment and choose different actions to generate rewards. Apart from reacting to random machine failures, the provided model enables collaborative robotic processes to further shorten the makespan and increase the utilization of industrial robots paired with machine tools.
In this study, we designed a control management mechanism for autonomous vehicles by utilizing the ownership Quality of Service (QoS) of Data Distribution Service (DDS) and implemented and experimentally verified it based on FastDDS. By setting priorities among the three control modes of manual control, Advanced Driver Assistance System (ADAS) control, and emergency control based on ownership strength, we prevented data conflicts and enabled fast and accurate control switching even in emergency situations. The experimental results show that the proposed system minimizes the control transition latency. Also, it effectively guarantees the message throughput and demonstrates its practical applicability in safety-centric systems such as autonomous driving. This research will provide future possibilities for extending ownership QoS in complex environments such as multi-vehicle cooperative driving and vehicle-infrastructure communication. It also aims to contribute to improving the reliability and efficiency of autonomous driving technologies.
For knee rehabilitation patients, it is difficult to maintain balance while walking, and completing a normal gait cycle is challenging. Therefore, we propose a lower limb exosuit tailored for the knee rehabilitation patients to assist in normal walking, and experimentally evaluated the assistance effect of the lower limb exosuit. We mimic the mechanism of the skeletal muscles near the knee joint, by contracting and relaxing the length of the flexible Bowden cable between the thigh and calf, generating assistive torque on the knee joint periodically. We use Inertial Measurement Unit (IMU) to measure lower limb motion parameters, identify gait cycle states, and calculate corresponding plantar pressure and knee joint torque for each phase. We compared surface electromyography (sEMG) signals of three muscle groups (biceps femoris (BF), quadriceps femoris (QF), and gastrocnemius (G)) under conditions of wearing and not wearing the exosuit. The experimental results show that compared to not wearing the exosuit, the mean normalized sEMG signals of the three muscle groups decreased by 4.61 %, 2.41 %, and 3.97% respectively while wearing it, the peak values decreased by 11.60%, 25.69%, and 14.99% respectively, with the quadriceps femoris receiving the highest degree of assistance.
An accurate and reliable image-based fruit detection system is essential for advancing agricultural tasks such as yield mapping and robotic harvesting. This paper benchmarks five state-of-the-art object detection frameworks on a merged dataset of 10 common fruit classes. The challenge of scarcity of high-quality fruit dataset detection of multiple classes of fruit is addressed. Data merging and augmentation techniques are used to increase data samples and capture variability. Multiple architectures are used to explore computational trade-offs of varying deep learning networks. Model accuracy and inference speed are evaluated for considerations of future deployment in an agricultural environment. This study offers guidance on model selection for fruit detection based on accuracy and latency requirements. Future research should evaluate latencies on resource-constrained devices and should prioritize improving accuracy while reducing overall complexity and inference speed for fruit detection tasks.
Advancements in multi-agent systems (MAS) have enabled swarm-based systems to perform decentralized decision-making and autonomous tasks. However, optimizing their performance while ensuring transparency and interpretability remains a challenge. This paper introduces a framework that combines Bayesian optimization with Explainable Artificial Intelligence (XAI) techniques to enhance both the efficiency and transparency of MAS swarms. The Bayesian optimization framework fine-tunes agent parameters to improve swarm metrics such as energy efficiency, task completion time, and coordination success. The experimental results show significant improvements: a 25% increase in the coordination success rate, a 15% increase in energy efficiency, and a 20 % reduction in task completion time. XAI techniques, including SHAP values, provide interpretable explanations for optimization decisions, improving user trust and understanding. This study demonstrates the efficacy of integrating Bayesian optimization with XAI to create transparent, efficient, and reliable MAS swarms. Future work should address scalability and implications in dynamic environments.
This paper presents a modular soft hand exoskeleton with an interactive interface designed to assist in rehabilitating individuals with spinal cord injuries experiencing hand spasticity. Following a Human-Centered Design (HCD) approach, the development process began with identifying and analyzing user needs, progressing through iterative phases of modeling, simulation, fabrication, and testing. The exoskeleton achieves a finger range of motion (ROM) of approximately 181.7 degrees, closely natural hand movements while maintaining a lightweight design of 96 grams. Its modular and portable architecture allows for adaptation to different hand sizes, enhancing accessibility. The system was evaluated through an interactive game developed for therapy sessions, demonstrating functionality and usability. Preliminary results demonstrate promising potential for hand rehabilitation, and future work will focus on refining the design and validating its effectiveness on a larger scale. This work offers a practical and adaptable solution for hand rehabilitation in individuals with spinal cord injuries, contributing to the ongoing progress in rehabilitation technologies.
This research introduces an advanced flight control system for optimizing autonomous aircraft performance, leveraging deep reinforcement learning (DRL) to address the complexities of nonlinear flight dynamics. Using a six-degree-of-freedom (6-DoF) rigid aircraft flight dynamics model, we develop a Deep Deterministic Policy Gradient (DDPG) controller tailored for waypoint navigation and attitude stabilization tasks. A custom reward framework and extensive hyperparameter tuning enable effective training within a high-fidelity MATLAB/Simulink environment, achieving high rewards and precise control. Although computationally intensive, the simulations demonstrate robust performance across diverse flight conditions, with potential for real-world applications and future extensions to multi-agent scenarios.
The growing interest in energy efficiency points the research for hydraulic systems to electro-hydraulic drives, specifically hydraulic actuators directly coupled to pumps driven by electric motors. This design eliminates functional energy dissipations and enables energy recovery but leads to inherently low damping. The consequence is that harmful payload oscillations deteriorate the position control's accuracy of the actuator. Therefore, raising the drive damping via pressure feedback is crucial and requires knowledge of the pressure in both actuator chambers. This paper investigates using pressure observation to that end (i.e., a virtual pressure sensor implemented via software replaces two physical pressure transducers). High-passed pressure feedback is introduced using a model-based design for setting up closed-loop position control. This approach is simulated numerically on a heavy-duty robotic manipulator. The results confirm that the payload oscillations are significantly reduced without affecting position tracking.
Efficient and robust design of legged robots for diverse terrains remains a critical focus in robotics research. To streamline this process we present varyMorph, an open-source tool that programmatically generates diverse legged robot models in both URDF and USDA formats. By randomizing properties such as leg count, morphology, body dimensions, and mass distribution, varyMorph creates hundreds or thousands of valid robot designs in few minutes. Users can seamlessly integrate these models into popular simulators (e.g., Isaac Sim/Lab, MuJoCo, Isaac Gym, RaiSim), accelerating large-scale experimentation in evolutionary robotics and reinforcement learning. Our results show near-linear scalability with respect to complexity and batch size, achieving rapid generation on both high-performance and standard desktop hardware. This high-throughput morphological exploration facilitates data-driven optimization and supports novel research on locomotion control, sim-to-real transfer, and co-evolutionary algorithms. We envision varyMorph as a complementary resource for grammar-based, latent-space, and gradient-driven design frameworks, reducing the barrier to discovering innovative and robust robot morphologies.
In this paper the PID-Fuzzy controller, which combine the robustness and versatility of controllers based on fuzzy logic, were implemented to achieve superior performance in regulating an aero-pendulum system. These controllers were specifically designed to enhance the regulation of the system by improving its stability, accuracy, and response to external disturbances. Through the integration of fuzzy logic, the control system was better equipped to handle the uncertain associated with the wind-driven operation of the system. In addition, a soft sensor was developed to estimate variables that could not be directly measured, utilizing both system measurements and process knowledge. This sensor improved the precision of the control system, enabling more accurate and reliable performance. The combination of these advanced control techniques resulted in a robust and effective control solution for the aero-pendulum system, making it well-suited for use in automatic control and mechatronics laboratory settings. The PID controller was developed using the Ziegler-Nichols tuning method, and each was subsequently enhanced through the incorporation of fuzzy logic. In addition, a comparative study was conducted to evaluate the performance of the controllers, highlighting the improvements achieved through the application of fuzzy logic. Finally, the performance of the designed controller were tested in real time aero-pendulum platform.
This paper presents the design and analysis of a smart solar panel cleaning robot, focusing on kinematic and dynamic modeling to enhance cleaning efficiency and structural integrity. The robot is designed to maximize photovoltaic panel efficiency by mitigating soiling effects caused by dust and debris accumulation. Its kinematic model ensures precise linear motion at a speed of 10 m/min, while the dynamic analysis optimizes torque and motor performance, ensuring smooth operation across large panel arrays. Stress analysis guided the selection of aluminum 1060 alloy for the robot's platform, offering a balance between lightweight properties, excellent corrosion resistance, and cost-effectiveness. Additionally, the robot employs a compressed-air cleaning mechanism, eliminating the need for water, making it ideal for arid regions. Powered by integrated solar panels and lithium-polymer batteries, the robot is energy-autonomous and capable of efficiently cleaning up to 100 meters of solar panels in approximately 10 minutes. This innovative design highlights advancements in kinematics, dynamics, and material engineering, contributing to the development of sustainable maintenance solutions for photovoltaic systems.
In a warehouse context, the goal of the research project is to prove the effectiveness of a fully automated, infrastructure-free guidance system for pallet transfer within warehouses. The means of transfer used is a fleet of Automatic Handling Units (AHUs). The goal is warehouse automation, which has become a priority for the largest players in the logistics sector. The growth in the volume of goods handled is a consequence of the globalisation of the world economy and the increasing demand for fast and efficient supply chain solutions. Thus, it forces warehouses to adapt to the growing complexity of operations. Competition between warehouses and the need to reduce lead times for goods handling require operators to increase overall productivity and reduce operating costs while guaranteeing maximum security. This article addresses the scientific challenge of automatic and optimized fleet management of AHUs. It involves proposing a routing system for the AHU fleet without deadlocks that allow regularly submitted missions to be carried out.
Aiming at the thermal error problem of CNC machine tools, the causes and changing rules of the thermal error of the feed axis and spindle were studied in depth. We proposed an efficient testing and modeling method for thermal error of feed axis and spindle. A thermal error compensation system was developed, which can realize the testing, modeling and real-time compensation of thermal error of feed axis and spindle. The real-time compensation results of CNC machine tools show that the thermal error was reduced by 80-90%.
Early detection of cancer is essential as it significantly improves treatment outcomes and survival rates by addressing the disease before it spreads to other organs. Many cancers, such as breast, cervical, and colorectal, are highly treatable in their initial stages. With the development of Artificial intelligence, nano technology, the engineers, doctors, and biologist are working on nanorobots for diagnosis and treatment of cancer. It can help detect cancer at an early stage. Early diagnosis often reduces the complexity and cost of treatment, alleviating the physical and financial burdens on patients. Nanorobots are at the forefront of advanced biomedical applications, offering potential breakthroughs in targeted drug delivery and precision diagnostics, particularly for cancer treatment. By integrating artificial intelligence and Deep Reinforcement Learning, these devices can autonomously navigate complex biological environments to detect cancer cells via biomarker gradients. This work explores the deployment of Deep Q-Learning to improve nanorobot decision-making, enabling efficient path optimization and obstacle navigation.
Target-tracing applications have attracted much interest in unmanned vehicles and mobile robots, especially where human access is limited or dangerous. Several computational techniques such as genetic algorithms, evolutionary computations, and neural networks have been explored to improve the entire system's control mechanism and efficiency. This paper addresses PSO as a high-performance method for enabling collaborative robotic search operations. It has many critical parameters influencing its performance, which are finely optimized using another PSO process to make the best use of this optimization technique. Experimentations set forth show that this indeed does much improve the efficiency and accuracy of searches carried out by robots. These validation experiments work on single-target and multi-target cases to present vivid effectiveness. The results emphasize the ability of PSO to contribute toward autonomous search and rescue operations and all other applications.
In our previous study, assist device equipped with headrest for overhead work was developed. In this paper, effect of the headrest for reducing burden on the neck when performing overhead work was investigated based on static analysis and dynamic analysis. In the static analysis, cervical spine load model was proposed, and the burden on the neck was evaluated statically. As results, it was verified that moment around C2/C3 intervertebral disc, and compressive force and shear force at the C2/C3 intervertebral disc were reduced due to the presence of the headrest. In the dynamic analysis, the motion of lowering arms when performing overhead work was created using the code of musculoskeletal model analysis software “AnyBody Modeling System”, and burden on the cervical spine was evaluated dynamically using AnyBody Modeling System. As results, it was verified that amount of muscle activity of sternocleidomastoid muscle, and compressive force and shear force at the C2/C3 intervertebral disc were reduced due to the presence of the headrest. Thus, usefulness of the headrest in the assist device for overhead work for reducing the burdens on the neck was demonstrated.
