This paper proposes a projection-based layer for constrained path planning, where task requirements are encoded as equality constraints defining an implicit manifold. The method is compatible with geometric planners that optimise a path inside a corridor represented as a sequence of convex safety regions, and returns a collision-free B-spline path, with user-prescribed degree and continuity, that is uniformly ɛ-adherent to the implicit task constraints. Corridor construction is biased toward task feasibility by projecting exploration seeds onto the manifold, so that each corridor element contains at least one feasible configuration. After corridor optimisation, the task constraints are enforced at the B-spline control points within a prescribed projection tolerance through a constrained closest-point problem that preserves corridor containment. Continuity is then recovered by a smoothing procedure based on alternating projections while maintaining the prescribed control-point tolerance. Uniform ɛ-adherence along the whole path is finally enforced by an outer loop that recomputes the knot vector with progressively more knots and repeats optimisation and projection until the prescribed tolerance is satisfied. The approach is validated through simulations and real-world experiments.
This study presents an integrated Autonomous Ground Vehicle (AGV) system for real-time pavement crack detection, quantification, and georeferenced mapping, enabling automated and high-precision infrastructure inspection. The system combines a hierarchical control architecture with an embedded deep learning perception framework. Crack detection is performed using a TensorRT-optimized YOLOv5s model deployed on an NVIDIA Jetson Xavier NX platform, operating at a high input resolution of 960 × 960 to enhance the detection of fine-scale crack features. The optimized model achieves a mean Average Precision (mAP@0.5) of 97.4% (with mAP@0.5:0.95 of 81.2%) while maintaining real-time performance at 28 frames per second on embedded hardware. To ensure accurate geospatial mapping, the AGV employs an Extended Kalman Filter (EKF)-based sensor fusion framework that integrates GNSS measurements with wheel odometry, achieving a localization Root Mean Square Error (RMSE) of 0.042 m. Detected crack features, including length, width, and severity, are quantified in metric units and georeferenced to global coordinates in real time. The system demonstrates reliable crack dimension estimation with sub-centimeter accuracy under typical conditions and stable embedded operation under real-world inspection conditions. The crack width estimation error ranges from approximately 1 mm under ideal conditions to over 4 mm in challenging scenarios such as shadow interference and surface contamination. The proposed AGV platform provides a fully autonomous, embedded, and georeferenced pavement inspection solution, offering high detection accuracy, precise localization, and real-time performance. This enables scalable, efficient, and quantitative pavement condition assessment for intelligent transportation infrastructure and proactive maintenance planning.
Broadband acoustic energy harvesting for fully implantable cochlear implants remains a significant challenge because conventional piezoelectric harvesters typically exhibit narrow operating bandwidths and limited electrical output under low-amplitude middle-ear vibrations. This study presents a metamaterial-inspired multichannel piezoelectric energy harvester that integrates hybrid chiral and non-chiral beam architectures with channel-specific optimized tip-mass geometries to enhance broadband electromechanical energy harvesting. Unlike conventional multichannel harvesters employing identical beam configurations, the proposed architecture enables each harvesting channel to utilize the beam topology and tip-mass geometry best suited to its target eigenfrequency. The proposed design was investigated through finite element modeling and systematic geometric optimization of the tip-mass configuration. The optimized harvester operates over a frequency range of approximately 250–4000 Hz, achieving a maximum average output voltage of 13.109 V, a cumulative maximum output power of approximately 1.405μW, and a maximum power density of approximately 5.65×105 W/m3. Compared with previously reported piezoelectric energy harvesters for cochlear implant applications, the proposed hybrid architecture provides substantially enhanced voltage generation, power output, power density, and broadband frequency coverage while satisfying the design constraints of implantable middle-ear devices. These findings demonstrate that the proposed channel-specific hybrid beam architecture provides an effective strategy for broadband acoustic energy harvesting and represents a promising energy harvesting component for supplementing the power budget or charging an intermediate energy storage unit in future fully implantable cochlear implant systems.
The growing demand for robotic systems in physical human–robot interaction (pHRI) has driven the need for advanced control strategies that ensure both task performance and human safety. This has led to more complex, sophisticated control strategies to overcome inherent parametric uncertainties and external disturbances encountered during pHRI tasks. Many recent strategies rely on learning-based methods, which, despite their excellent results, are often limited by computational demands rather than control design matters. These learning-based approaches require vast amounts of data and extensive training times. As an alternative, indirect adaptive control offers a more efficient solution by enabling real-time identification and compensation for changes in the robotic system during its operation. While indirect adaptive control has emerged as a promising approach for handling time-varying dynamics, its real-time implementation can be computationally intensive, especially with complex models or when iterative identification is used. To address this, the present work proposes a novel, lightweight surrogate model-based indirect adaptive control strategy that combines radial basis function regression with a metaheuristic optimization algorithm. The proposed strategy iteratively reconfigures the robot’s null space when external forces are detected along its kinematic chain, aiming to minimize them. At the same time, the parameters of a Cartesian impedance controller are dynamically updated to preserve the end-effector performance. The proposed strategy is validated using a Franka EMIKA Panda robot manipulator, addressing regulation and path-tracking control problems, as well as in a constrained task. The outcomes show the proposal’s capability to reconfigure the null-space while maintaining the end-effector position and orientation squared errors within the 10−3 and 10−1 ranges, respectively.
This study presents a new distributed piezoelectric self-sensing technique for pressure estimation in a smart structure under static and quasi-static conditions. The system consists of five piezoelectric elements designed to detect localized and distributed pressures from objects of various shapes. A charge model improves accuracy by incorporating feedthrough, hysteresis, and creep effects. A custom electronic circuit is designed to interface with the piezoelectric structure, providing charge amplification and compensation for feedthrough effects. In addition, an estimator is implemented to perform real-time compensation for nonlinear behavior and to calculate the applied pressure based on the effective output voltage and the identified self-sensing parameters. The self-sensing parameters are experimentally validated to ensure accurate compensation for electrical parameters, hysteresis, and creep effects. The system is further tested using known pressures and various shaped objects, including a circular disk, a rectangular ruler, and a mini-book. These configurations demonstrate the system's ability to discriminate between localized and uniformly distributed pressure profiles effectively. The results demonstrate a strong correlation between estimated and applied pressures, with errors remaining below 7 % under optimized compensation, greatly surpassing the performance of a conventional distributed piezoelectric structure used only as a sensor (Vin=0V). Integrating the charge model, custom circuitry, and estimator highlights the robustness of the proposed approach and provides a reliable solution for autonomous pressure mapping in practical applications.
Industrial robots suffer from limited absolute positioning accuracy, which restricts their applications in precision manufacturing. To address this limitation, this paper proposes a high-order joint-dependent kinematic error modeling and compensation method that explicitly accounts for joint flexibility induced by the robot’s self-weight and external payloads. Unlike conventional calibration models that assume constant kinematic errors, the proposed approach incorporates flexibility-related parameters into a high-order joint-dependent error formulation, enabling more accurate representation of configuration- and load-dependent positioning errors. In addition, a hybrid sampling strategy is developed to optimize the selection of measurement configurations for parameter identification. Joint-related geometric error parameters and flexibility parameters associated with self-weight and external payloads are identified, and the resulting model is applied for positioning error compensation. Experimental results demonstrate that the proposed method significantly improves the robot’s absolute positioning accuracy. Specifically, the maximum positioning error is reduced from 4.197 mm to 0.115 mm, while the average positioning error decreases from 1.405 mm to 0.043 mm. Furthermore, comparative experiments under different external payloads show that the proposed method consistently achieves the lowest root mean square error (RMSE) among several existing error models, demonstrating superior generalization capability.
This work presents an adaptive control strategy for a massless flexible-link robot with two lumped masses. Due to their sensitivity to parameter variations, flexible-link manipulators are prone to instability when controller parameters are not accurately tuned. To address this issue, a nested-loop adaptive control architecture is proposed, where the inner loop controls the motor angle, and the outer loop regulates the tip position through the base moment. A novel algebraic identification algorithm is incorporated for real-time parameter estimation, providing rapid convergence and strong robustness against strain-gauge disturbances. The estimated parameters are then used to tune the controller gains adaptively. Experimental results validate the effectiveness of the proposed approach, demonstrating significant improvements in speed, accuracy, and robustness. The method also shows strong potential for extension to flexible robotic arms with distributed mass and multiple vibration modes.
The development of precision motion stages with low moving mass and excellent control capability is critical for a wide range of high-tech applications. While over-actuation strategies have demonstrated excellent performance, existing approaches rely on heuristic structural designs and typically only regulate the fundamental flexible mode, which can limit system performance and generalizability. To address these issues, this paper proposes a novel topology-optimization (TO)-guided hardware-control co-design framework for lightweight over-actuated precision motion systems. The proposed approach utilizes TO for structural concept exploration, followed by sequential stage structure shape and feedback controller optimization. The framework is evaluated by the design optimization for a magnetically levitated precision stage prototype, evaluating schemes with up to three actively controlled flexible modes. Simulation results show that the TO-guided designs can outperform manual baselines, achieving a 100 Hz bandwidth in rigid-body DOFs alongside a more than 20% reduction in moving mass. Finally, an experimental prototype controlling three flexible modes using a combination of electromagnetic actuators and piezo actuators was fabricated and tested, successfully verifying the targeted 100 Hz rigid-body bandwidth while identifying critical coupling challenges in over-actuated systems using distributed piezo actuators.
Stair-climbing capability is essential for mobile robots operating in real-world industrial environments. Open-riser stairs, widely used in facilities such as oil and gas plants due to their low cost and high durability, present significant challenges for conventional wheeled robots because of limited contact support and increased risk of instability. This study proposes a wheeled robotic platform equipped with four driving wheels, a rear-mounted linear actuator, and an auxiliary wheel to improve climbing performance on open-riser stairs. The actuator provides additional lifting force during ascent, assisting the front wheels in overcoming step edges while regulating the robot’s pitch to prevent backward tipping. A thorough static force analysis is developed to investigate the climbing conditions and to provide theoretical guidance for the mechanical design. Based on the analysis, an autonomous control strategy is implemented using real-time pitch feedback and a finite-state-machine-based switching logic to coordinate wheel motion and actuator deployment. A prototype system is built and experimentally evaluated on open-riser staircases. The results demonstrate reliable autonomous stair climbing and show that the proposed mechanism improves the adaptability of conventional wheeled robots to open-riser stairs environments.
The frequent lifting and transfer of bedridden patients impose a considerable physical burden on nursing staff. To reduce this burden, this paper presents a novel four-arm robotic system for patient transfer, designed to achieve safe, efficient, and coordinated handling in healthcare environments. The paper focuses on the robotic platform and provides a detailed description of its mechanical architecture, control system, and practical application in transfer tasks between healthcare facilities such as beds and wheelchairs. To realize coordinated motion among the four manipulators, a trajectory optimization framework based on nonlinear model predictive control is developed to generate smooth and collision-free motions. Experimental validation in a representative patient transfer task demonstrates the feasibility, coordination capability, and practical effectiveness of the proposed robotic system.
The increasing use of robotic platforms in hazardous and magnetically complex environments, such as underground research facilities at CERN, demands a detailed understanding of how strong external magnetic fields affect electromechanical actuation. Quadruped robots equipped with electromagnetic (EM) motors are particularly vulnerable, as their actuators interact directly with ambient magnetic fluxes, generating additional mechanical forces and moments that can compromise stability, trajectory tracking and overall operational safety. This work presents a modeling framework for estimating the equivalent magnetic moment of an EM motor subjected to high-intensity magnetic fields. The proposed approach combines numerical simulation datasets with an analytical dipole-based formulation to derive a computationally efficient model capable of predicting magnetic-induced mechanical disturbances in real/time. The model accounts for the dependence of the magnetic moment on both the external field intensity and the motor’s orientation relative to the field lines, enabling the control system to anticipate disturbance torques and implement appropriate compensation strategies. The accuracy and reliability of the model were validated experimentally by rotating a Unitree GO-M8010-6 motor within controlled magnetic fields and comparing the measured torques with the model predictions. The results demonstrate close agreement in both trend and amplitude, confirming that the proposed formulation can provide the level of fidelity required for real-time control applications in magnetically harsh environments.
This work introduces a new hybrid Prandtl–Ishlinskii (PI) model. It uses a classical PI (CPI) model augmented with Long Short-Term Memory (LSTM) networks to represent the rate-dependent non-linearity, amplitude-dependent and asymmetrical hysteresis. The proposed PI-LSTM hysteresis model is then cascaded with a linear dynamics to create an extended Hammerstein approximation. The entire model is used to simulate and control hysteresis in shape memory alloys (SMAs). The hybrid formulation captures rate- and amplitude-dependent hysteresis by combining the interpretability of the CPI model with the adaptability of recurrent neural networks. Linear dynamic and non-linearity characteristics are identified through a rate- and amplitude-dependent system identification procedure and used to train the hybrid model. Experimental validation confirms that the proposed model achieves higher accuracy than the CPI model in predicting complex hysteresis behaviors. An inverse-multiplicative structure controller built on the hybrid model further enhances displacement tracking performance. Results highlight the efficacy of combining phenomenological and data-driven approaches for accurate modeling and control of SMAs for exoskeleton actuation.
Micro aerial vehicles (MAVs), and quadrotors in particular, have become essential platforms for research and applications requiring agile and reliable flight. However, developing accurate models and well-tuned controllers for quadrotors is a significant challenge. This task is further complicated by mechanical coupling to external structures, such as gimbal testbeds, which introduce dynamic interactions that can degrade control performance. To address this challenge this paper proposes a relay-based approach for the identification of linear models and the autotuning of PID controllers for a quadrotor attitude and rate dynamics, using a Crazyflie 2.1 MAV mounted on a gimbal structure. These models are used to fine-tune the PID through the SIMC and AMIGO tuning PID rules. An event-based control strategy with an adaptive threshold is also implemented to optimize computational efficiency while maintaining control performance achieving a 30.56% reduction in CPU load compared to periodic execution. Experimental results demonstrate that the method significantly improves performance compared to the default controller, which fails under the additional inertia introduced by the gimbal. The study highlights the potential of relay-based autotuning for MAVs and establishes a foundation for future benchmarks and free-flight validation.
Although sample-based model predictive control (MPC), such as model predictive path integral control (MPPI), are well suited to manage complex tasks like tracking a teleoperated robot while maintaining constraints and avoiding obstacles, it can be challenging to design the MPPI input sampling to achieve precision tracking. The main contribution of this work is to enable precision tracking with MPPI-type sampling methods by (i) sampling the system's reference outputs and (ii) using inversion based control to correct for system dynamics. An advantage of the proposed reference-output-sampled MPPI (oMPPI) is that the selected reference output's sample distribution can reflect the desired output of the system, such as the trajectory of the teleoperated robot in the active vision application. The proposed oMPPI is applied to a crane-robot active vision system for confined space inspection during aircraft wing manufacturing, enabling a teleoperated manipulator to navigate around in-wing structures. Teleoperation experiments show that oMPPI sampling increases tracking precision of a teleoperated manipulator by 22% and reduces camera oscillations by 65% when compared to MPPI sampling without inversion.
Energy consumption and reconfigurability are critical performance metrics for robotic systems, and their synergistic optimization is essential for enhancing operational versatility. This paper proposes an innovative lowenergy modular robot comprising a pitch joint module, a yaw joint module and an end-effector. The pitch joint module utilizes a parallelogram mechanism; taking a motion range of +40 degrees as a representative case in this paper, a gravity compensation unit is integrated to counterbalance the arm's weight, thereby significantly minimizing joint driving torques. Based on a planetary gear train design, the yaw joint module achieves a yaw angle of +180 degrees, which effectively extends the robot's workspace. The gravitational compensation mechanism for pitch unit is designed with high-performance thermoplastic elastomer (TPE) strands. By precisely tuning the internal geometric installation parameter, the mechanism provides a tunable compensation range to counteract the varying gravitational moments of different module assemblies, ensuring adaptability to diverse load conditions. The performance of this mechanism was evaluated through detailed energy analysis and robot pick-andplace experiments. The experimental results demonstrate significant energy savings: The prototype is capable of executing grasping tasks, the compensation unit reduces pitch module energy consumption by 83 % and 82 %, respectively achieves a 27.7 % decrease in overall system energy consumption during material handling, thereby validating the robot's low-power and modular characteristics.
This paper presents a gain-design methodology for a Proportional-Integral-Derivative (PID)-like sliding-mode controller applied to the elevation subsystem of a three-degree-of-freedom (3-DOF) helicopter prototype. The dominant elevation dynamics are modeled as a double integrator with static gain, while parasitic effects are represented by a transport delay. The proposed methodology comprises two steps. First, the Robust Feedback Self-Oscillation Test (RFSOT) is employed to identify the magnitude of the parasitic delay. Second, a systematic gain-tuning strategy is developed, based on the describing function approach, to minimize either the amplitude of the fundamental chattering harmonic, the root-mean-square (RMS) value of the control signal or the average power needed to maintain the trajectories of the system in a real sliding mode. The effectiveness of the proposed approach is validated through numerical simulations and real-time experiments conducted on the elevation subsystem of a 3-DOF helicopter laboratory platform.
Hydraulic excavators are widely used in construction tasks such as levelling and grading, which require high precision and repeatability. Automating these operations is challenging due to the strong nonlinearities, coupling effects, and variability of hydraulic systems. In this work, we present a data-driven Nonlinear Model Predictive Control (NMPC) framework to autonomous grading operations of hydraulic excavators. The system dynamics are modelled using Local Linear Neuro-Fuzzy Models identified from experimental input-output data, enabling a flexible representation without requiring detailed physical modelling. Two NMPC formulations are developed and compared: a trajectory-tracking approach and a path-following approach, the latter allowing for adaptive timing along the desired path. The proposed methods are experimentally validated on a full-scale JCB Hydradig 110 W excavator performing levelling and sloping tasks. The results show that both NMPC approaches achieve accurate tracking performance under realistic operating conditions. Moreover, the path-following formulation demonstrates improved robustness in scenarios where flexibility in execution speed is beneficial. A comparison with a baseline data-driven controller highlights the effectiveness of the proposed approach.
