ABSTRACT. Intelligent vehicles have received considerable attention in recent years. Despite relevant advances in the last two decades, there are still challenges and open questions to be addressed in this area. This talk presents research results on intelligent vehicles that range from basic software components to algorithms for intention prediction and traffic negotiation. 

ABSTRACT. This paper presents a control framework for the retrieval operation of an inspection crawler from a pipeline, using an UAV (Unmanned Aerial Vehicle) with a cable. The inspection crawler inspects the pipeline until it is retrieved using the UAV with a cable. In order to create this control framework, the task has been divided in two different phases: the tethered UAV phase and the flight of the UAV with a suspended load. For each phase, we study the dynamic and we propose a specific controller. The control framework is composed of a controller obtained using the feedback linearization technique for the tethered phase, while the controller of the second phase is based on an IDA-PBC (Interconnection and Damping Assignment-Passivity Based Control) controller. The proposed control framework is validated through simulations in Matlab-Simulink, showing that the UAV can properly recover the inspection crawler from the pipeline and stabilizes with the crawler as a suspended load at a new position.

ABSTRACT. Similarly to the Control Lyapunov Functions (CLFs), whose objective is to achieve the stability of a system, Control Barrier Functions (CBFs) aim to achieve the safety of a system. They serve as a safety filter that guarantees that the system remains in a defined safety region. This article aims at presenting an introductory overview of the theoretical framework of CBFs and of their application. For doing this, we apply the CBFs framework as a safety filter for obstacle avoidance in the XY movement of a multirotor. The safety filter is designed using two different multirotor models, a single integrator model without considering the inertia of the system and a first-order model at the velocity loop, which consider the inertia of the system. The proposed safety framework is validated with simulations in Matlab and experiments with indoor flights. The results show the importance of correctly modeling the system.

ABSTRACT. The cooperation between Unmanned Aerial Vehicles (UAVs) and Unmanned Ground Vehicles (UGVs) has brought new perspectives and effectiveness to production and monitoring processes. In this sense, tracking moving targets in heterogeneous systems involves coordination, formation, and positioning systems between UGVs and UAVs. This ar- ticle presents a Proportional-Integral-Derivative (PID) control strategy for tracking moving target operations, considering an operating envi- ronment between a multirotor UAV and an indoor UGV. Different PID architectures are developed and compared to each other in the Gazebo simulator, whose objective is to analyze the control performance of the UAV when used to track the ground robot based on the identification of the ArUco fiducial marker. Computer vision techniques based on the Robot Operating System (ROS) are integrated into the UAV’s tracking system to provide a visual reference for the aircraft’s navigation system. The results of this study indicate that the PD, Cascade and Parallel controllers showed similar performance in both trajectories tested, with the Parallel controller showing a slight advantage in terms of mean error and standard deviation, suggesting its suitability for applications that prioritize precision and stability.

ABSTRACT. The ability for a robot to be able to construct a map of the environment and recognize its position on it was one of the biggest developments in robotics. Simultaneous localization and mapping (SLAM) framework builds onto the perception of the robot, giving it the possibility to online calculate its trajectory and avoid obstacles. Moreover, the continuous development of processing units has given the possibility for previously hardware exhausting solutions to be considered as an option for the localization and mapping problem. With this in mind, this work is focused on developing a SLAM solution for a 6 DoF vehicle operating on a 3-D environment using moving horizon estimation (MHE). Throughout the paper it is tested the applicability of the proposed solution in a simulation environment of two loop square-shaped corridors with stationary landmarks, whilst comparing the obtained results with another probabilistic approach, the EKF, which is commonly used but loses stability on extremely nonlinear dynamics. Each of the algorithms is simulated in MATLAB.

ABSTRACT. The utilization of unmanned vehicles for specialized tasks has gained significant attention in both military and civilian domains. This article explores the application of commercial unmanned aerial vehicles (UAVs) for reconnaissance purposes, specifically to verify autonomous driving missions assigned to the developed TAERO manned-unmanned vehicle in field operations. The paper introduces the TAERO vehicle, highlighting its functionality and capabilities for unmanned missions. The architecture of the unmanned ground vehicle (UGV) system is discussed, taking into consideration the autonomy subsystem and used location data. The limitations associated with terrain and potential obstacles are addressed, as well as the importance of acquiring accurate terrain information for successful autonomous operation. The solution proposed in our study involves the use of a commercially available UAV applied to the visual tracking of potential targets in an engagement scenario. Details related to flight route planning system, geolocation, target tracking, and data transmission between robotic platforms are discussed and presented in this work. The acquired real-time data plays a crucial role in confirming the mission, making necessary adjustments, or altering the planned route. The UAV platform, known for its maneuverability and operational capabilities, can operate ahead as a reconnaissance element, improving the overall reconnaissance capabilities of the system. Upon completion of the mission, the UAV can return to the base or land on a moving vehicle platform. The authors proposed integration of a UAV that significantly enhances the autonomous mode capabilities of unmanned ground platform, improving operation in unknown environment during special mission.

ABSTRACT. The development cadence of deep learning algorithms is extremely high, with the development of autonomous navigation systems and especially autonomous vehicles in the forefront. However, growing environmental awareness has directed efforts toward the development autonomous system for the maintenance and preservation of forested areas. Unlike urban areas, available datasets on these environments are scarce and incomplete. In addition, the difficult recording activities, together with the cumbersome labeling process, result in a high rate of mislabeling. Given the success of using synthetic data in model training, this research provides an approach that overcomes these limitations. The SynPhoRest system can extract photorealistic synthetic data in the form of RGB images, semantic segmentation maps and depth maps from procedurally generated virtual forest environments. Furthermore, the system supports both manual and automated extraction and can generate 100 sets of data in approximately one hour.

ABSTRACT. Obstacle detection is an essential component for autonomous vehicles to navigate safely. To overcome some drawbacks of 2D object detection, many recent methods have been proposed to cope with 3D detection using LiDAR sensors. This sensor provides depth and more representative spatial and geometric information to the models, allowing the estimation of 3D bounding boxes with orientation around the objects of interest. However, the complexity of 3D annotation, in supervised object detection, imposes significant challenges for this task. To address the annotation problem, some studies have explored data augmentation techniques for 3D point clouds, but, not all augmentation methods yield positive impacts on model performance. Therefore, this paper presents an in-depth evaluation of global data augmentation techniques, specifically focusing on geometric transformation and noise-based methods for 3D object detection. The results reported in this paper, achieved on the KITTI dataset, showed a relevant difference in some geometric operations, and the importance of noise-based methods.

ABSTRACT. Recent advancements in deep learning have provided powerful tools for intelligent vehicle tasks, particularly in the field of perception. However, achieving real-time performance with low power consumption remains a challenge included in the hot research topic known as green deep learning. In this paper, we present a comparative analysis of various YOLOv5 weights trained on the KITTI and SHIFT datasets using two platforms with different power consumption profiles: the NVIDIA Jetson AGX Xavier and a desktop computer equipped with a NVIDIA GTX 1080 Ti. Our analysis focuses on the average inference time and precision metrics for road objects detection, a key task for intelligent vehicles. Additionally, we apply TensorRT to optimize and accelerate the architecture on both platforms, resulting in significant speed improvements, particularly on the low-power Jetson AGX Xavier (30W). Our ultimate goal is to implement our whole autonomous driving architecture on several Jetson AGX Xaviers connected to a PC where the hyper-realistic CARLA simulator, is replicating the real-world autonomous vehicle environment. We obtain compelling validation results on KITTI and CARLA, achieving real-time performance on a lightweight Jetson AGX Xavier with a powerful object detector such as YOLOv5m.

ABSTRACT. Data collection and scene understanding have become crucial tasks in the development of intelligent vehicles, particularly in the context of autonomous driving. Deep Learning (DL) and transformer-based architectures have emerged as the preferred methods for object detection and segmentation tasks. The combination of advanced hardware and DL has significantly advanced scene understanding and object detection in recent years. However, DL-based methods often require extensive training with diverse data, posing challenges in terms of data availability and labeling. To address this problem, techniques such as transfer learning and data augmentation have been adopted. Simulators like CARLA have gained popularity in the autonomous driving domain, enabling the evaluation of architectures in realistic environments before real-world deployment. Synthetic data generated by simulators offers several advantages, including cost-effectiveness, access to diverse scenarios, and the ability to generate accurate ground truth annotations. This paper focuses on evaluating the performance and domain adaptation of a 3D object detection pipeline based on depth estimation using a stereo camera in CARLA simulator.

ABSTRACT. Detection and avoidance of dynamic obstacles is an integral part of social robot navigation. Reliable human detection depends on camera identification, which can be achieved only using computationally expensive algorithms running on a Graphical Processing Unit (GPU). The process is time consuming, causing latency and it cannot be run on low-end systems. Human detection and tracking also requires lidar data fusion to ensure proper localization. In this work, we propose a detection strategy that allows the fusion system to run with lower camera frame rates, and hence decrease latency and computational requirements considerably. We show the effectiveness of the proposed approach in simulation.

ABSTRACT. Simulation has become a fundamental practice for development in Robotics Industry and is reaching new heights with trends such as Digital Twins. But how can the power of simulation be combined with the impressive models offered by Artificial Intelligence?

In this session, we will present four growing trends in Robotics Industry that combine AI and Simulation. We will go beyond the mere use of AI models in Simulink, showing an unknown potential in many cases.

ABSTRACT. The past few decades have witnessed a significant research effort in the field of Lyapunov model based control design. In parallel, optimal control and optimization model based design have also expanded their range of applications, and nowadays, receding horizon approaches can be considered a mature field for particular classes of control systems.

In this talk, I will argue that Lyapunov based techniques play an important role for analysis of model based optimization methodologies and moreover, both approaches can be combined for control design resulting in powerful frameworks with formal guarantees of robustness, stability, performance, and safety. Illustrative examples in the area of motion control of autonomous robotic vehicles will be presented for Autonomous Underwater Vehicles (AUVs), Autonomous Surface Vehicles (ASVs) and Unmanned Aerial Vehicles (UAVs).

ABSTRACT. To evaluate the performance of data-driven control methods applied to Unmanned Aerial Vehicles (UAVs), this work addresses the implementation of these strategies, particularly the Data-enabled Predictive Control (DeePC) algorithm. This strategy computes optimal controls for unknown systems through real-time output feedback using a receding horizon implementation. Moreover, this research investigates the influence of different hyperparameters on the DeePC’s performance and conducts a realistic comparison between this method and two model-based control approaches: Linear Quadratic Regulator (LQR) and Model Predictive Control (MPC). The simulation results validate the applicability of DeePC algorithm and highlight its superior robustness to system degradation and yaw calibration errors. However, it is less suitable for complex nonlinear systems subject to aggressive trajectories.

ABSTRACT. This paper addresses the use of model predictive control (MPC) for pursuit-evasion games, where a shuttle drone aims at capturing a remotely piloted target drone. Several models for shuttle and target vehicles are developed, considering different degrees of complexity and reliance in inner-loop controllers. Based on these models MPC strategies are defined for each vehicle, for joint tracking operation, as well as for differential pursuit-evasion games between shuttle and target vehicles. Lastly the human-in-the-loop is added to the model and respective MPC algorithms of the fixed-wing target. Simulation results are provided, showing several scenarios where each vehicle has the advantage in terms of physical capabilities, or the disadvantage of being remotely-piloted.

ABSTRACT. Unmanned Aerial Vehicles (UAVs) have emerged as valuable tools in precision agriculture due to their ability to provide timely and detailed information over large agricultural areas. In this sense, this work aims to evaluate the semi-autonomous navigation capacity of a multirotor UAV when applied in the field of precision agriculture. For this, a small aircraft is used to identify and track a set of fiducial markers (Ar\_Track\_Alvar) in an environment that simulates inspections of insect traps in olive groves. The purpose of this marker is to provide a visual reference point for the drone's navigation system. Once the Ar\_Track\_Alvar marker is detected, the robot will receive navigation information based on the marker's position to approach the specific trap. The experimental setup evaluated the computer vision algorithm applied to the UAV to make it recognize the Ar\_Track\_Alvar marker and then reach the trap efficiently. Experimental tests were conducted in a indoor and outdoor environment using DJI Tello. The results demonstrated the feasibility of applying these fiducial markers as a solution for the UAV's navigation in this proposed scenario.

ABSTRACT. This paper presents a framework for simulating the visual tracking of low orbit space platforms such as satellites to be captured by a space manipulator system intended to conduct assembly, maintenance, or deorbiting operations. The video stream from the International Space Station, publicly available, is used as background for an animated overlay of the target platform moving within the field of view of the camera. A real-synthetic video is generated in Blender in different daytime observation conditions, allowing also the introduction of light sources for simulating the sun or its reflection of Earth' surface. The Continuously Adaptive Mean-Shift (CAMShift) algorithm is evaluated considering a satellite platform as tracked target with different approaching trajectories, illumination conditions, and background. A border detection stage is introduced to enhance the detection phase in dark scenes, exploiting the linear segments typical of satellites construction and solar panel arrays. The performance of the algorithm is evaluated in three different background conditions, comparing the execution time of C++ and Python implementations.

ABSTRACT. Driver distraction recognition plays a fundamental role in road safety. In this paper we present a modular architecture based on the fusion of key points and object detection for predicting driver's actions. From multi-camera infrared recordings, we will temporarily detect among a variety of actions that lead to distractions. Our system detects objects of interest and extracts key points from the driver. They are merged by generating features that relate them and processed with a ML-based classification algorithm. Finally, filters are applied to reduce bounces and add temporal context to the detections. Our proposal has been validated on two state-of-the-art datasets for driving distractions. Through several experiments we show that fusion substantially improves related action inference and improves domain adaptation. In addition, our framework is lightweight, explainable and has a low latency as it performs frame-by-frame inference. The modularity of the network allows us to upgrade parts independently or eliminate a camera without having to modify the entire network.

ABSTRACT. Egocentric action recognition is the ability of identifying hu- man actions from videos taken in a first-person perspective. In this thesis, the context of human-robot interaction is considered, where the primary objective is to comprehend the actions and behaviors of individuals to- wards the observer. However, video data of human interactions taken from a third-person perspective (exocentric view) is significantly more abundant, compared to the egocentric perspective. Thus, we propose an approach to train action recognition models using exocentric datasets, that can be used in run-time to classify egocentric data. Our approach relies on a feature space, based on skeleton pose information, strategi- cally designed to be consistent across exocentric and egocentric domains. The results show that models trained on exocentric data can be used on an egocentric view without any fine-tuning or adaptation. The system proved to generalise well in the unseen first-person domain, achieving an accuracy of 97% on a custom egocentric dataset. This score was obtained after combining several independently trained models into an ensemble, recognising actions in a voting mechanism. Using ensembles allowed for higher accuracy scores, and mitigated the variability between perfor- mances of different models, proving to be an advantageous alternative to using a single model. This method unveils new possibilities for advanc- ing research and developing more efficient models for first-person action recognition.

ABSTRACT. Physical Human-Robot Interaction (pHRI) is essential for robots that establish contact with humans, like assistive or rescue robots. pHRI demands hard safety and compliance constraints. Tactile sensing is vital for pHRI, as vision and/or range sensors alone can not cope with constant occlusions while grasping. Measuring pressure on the grasping surface is crucial to avoid injury, predict user intent and successfully perform collaborative movements. This work presents a new three-fingered gripper equipped with novel tactile sense capability via arrays of finger pressure sensors. Experiments gripping a human arm on the distal and proximal surfaces of the gripper fingers have been performed to assess sensing performance in the normal and tangential components. Results prove that a good estimation of interaction between human and robot is achieved in both normal and tangential movements.

ABSTRACT. Lower Extremity Fugl-Meyer Assessment (FMA-LE) is recommend-ed as the primary outcome for assessing motor function in post-stroke popula-tion. However, the subjectivity, dependency on professional experience, and time-consuming visual inspection by healthcare professionals limit the use of FMA-LE in clinical practice. Contrarily to clinical scales, sensor-based assess-ments can automatically provide objective measurements of motor function. This work advances literature by evaluating the Spearman correlation between the FMA-LE clinical scores and both spatiotemporal and electromyographic (EMG) measures, acquired during different mobility walking tasks (self-selected speed, maximum speed, maximum cadence, maximum step length, and maximum step height). Data were extracted from ARRA dataset, including 27 post-stroke participants. The results showed that step length (0.44 ≤ r ≤ 0.60), stride time (-0.48 ≤ r ≤ -0.40), and cadence (0.40 ≤ r ≤ 0.46) spatiotemporal measures, and peak power frequency (PKF) EMG measure of gluteus medius (r = 0.42), lateral hamstring (0.40 ≤ r ≤ 0.46), and vastus medialis (0.42 ≤ r ≤ 0.45) muscles revealed significant strong correlations in multiple walking tasks. Overall, spatiotemporal measures presented higher correlations with FMA-LE than EMG measures. These findings are promising for future research to devel-op artificial intelligence methods to estimate the Lower FMA clinical scores for motor assessment, maximizing its use in clinical practice.

ABSTRACT. Biofeedback is a promising tool to complement conventional physi-cal therapy by fostering active participation of neurologically impaired patients during treatment. This work aims at a user-centered design and usability as-sessment for different age groups of a novel wearable augmented reality (AR) application composed of a multimodal sensor network and corresponding con-trol strategies for personalized biofeedback during gait training. The proposed solution includes wearable AR glasses that deliver visual cues controlled in re-al-time according to mediolateral center of mass position, sagittal ankle angle, or tibialis anterior muscle activity from inertial and EMG sensors. Control strat-egies include positive and negative reinforcement conditions and are based on the user’s performance by comparing real-time sensor data with an automatical user-personalized threshold. The proposed solution allows ambulatory practice on daily scenarios, physiotherapists' involvement through a laptop screen, and contributes to further benchmark biofeedback regarding the type of sensor. Alt-hough old healthy adults with low academic degrees have a preference for guidance from an expert person, excellent usability scores (SUS scores: 81.25-96.87) were achieved with young and middle-aged healthy adults and one neu-rologically impaired patient.

ABSTRACT. This paper presents the modeling, control, and simulation of a flapping wing aerial platform equipped with a perching-launching mechanism and a two-degree-of-freedom (DoF) robotic arm intended to conduct manipulation tasks in outdoor scenarios which provide linear support structures for perching, considering as representative application example the contact inspection on power lines. The operation can be divided in four phases: 1) gliding towards the line, 2) perching, 3) manipulation, and 4) launching for flying again. A model of the system is derived and particularized for each phase, relying on numerical methods for simulating the perching and launching phases, following the Lagrange formulation for the equivalent planar manipulator once perched. A state-dependant Riccati equation (SDRE) controller is implemented on the robotic arm to conduct the manipulation task, taking into account that the passive pitch rotation of the base introduces an underactuated dynamics when perched. The model of the system, implemented in Simscape, is validated in simulation.

ABSTRACT. This paper presents an adaptive solution for the control of a quadrotor operating close to the ceiling. The vicinity alters the flow resulting on a variation on the effective thrust. A force factor is estimated in real time using an adaptive mechanism. The full controller is proven to have global asymptotic stability with zero tracking error using Lyapunov theory. The resulting system is validated in simulation with approaching trajectories to the ceiling, and its robustness is shown for mass variations of up to 20%.

ABSTRACT. This paper proposes a vector field guidance law for a team of unmanned aircrafts (UAs) to track a stationary target in a circular path at a fixed stand-off radius from the position of the target. A phasing Lyapunov function is also given and using the Lyapunov stability analysis, it is ensured that the two UAs will maintain a constant phase difference between them once they converge to the stand-off radius. To counter the effect of a constant background wind, the vector guidance law is modified using a variable scaling factor. During the transient phase i.e., before converging to the stand-off radius, the possibility of having a collision between the two UAs is avoided by using a Control Barrier Function (CBF) approach to generate a constraint based input control functions which avoids the collision between two UAs. We also propose an algorithm each for a periodic communication as well as event-based communication to reduce the frequency of the information exchange between UAs. Simulation results for the cases of constant communication between UAs, periodic communication and an event-based communication is presented which verifies the efficacy of the proposed methodology.

ABSTRACT. Locating odour sources with mobile robots is a difficult task that can be applied to locating the sources of pollutants, concealed explosives or victims in disaster scenarios. The existing approaches for locating odour sources can be divided between those that simply seek to reach the chemical source, and those that use gas dispersion models to estimate its location. One of the most popular source estimation approaches is Infotaxis, which has been shown to have great sensitivity to the parameters of its gas distribution model. In this paper, we compare two evolutionary approaches for automatically selecting the values for these parameters along with a Genetic Programming approach for evolving human-readable source-seeking strategies. The comparisons are carried out in three simulated environments with different chemical plumes and the results show that the parameters that best fit the environment do not always lead to the highest performance. Also, depending on the scenario, the tree-based search strategies are able to perform equivalently to Infotaxis, at a lesser computational cost.

ABSTRACT. This work deals with the problem of detecting and localizing methane emission sources in open spaces with a remote gas detector (TDLAS) equipped on a mobile robot. To alleviate the long inspection times of traditional approaches that use the ground as the natural reflector, in this work we analyze the feasibility of a leader-follower formation, where one robot will carry the remote gas detector horizontally, parallel to the ground, and the second robot will act as an artificial reflector. We present a visual tracking approach for the relative pose estimation of both mobile platforms to extend the measurement range up to 10 m. Results in a 70 m2 experimental area demonstrate that this approach is effective for a fast location of methane gas sources.

ABSTRACT. Cognitive search strategies are highly suitable for Odour Source Localisation (OSL) under turbulent flow conditions where continuous cues are not always available. While reliable, these methods have a known tendency to shift the weight of the search towards exploration to improve information gains. A balanced trade-off between exploration and exploitation (EE) is crucial since it can directly influence both the efficiency and reliability of the mission. However, its understanding from a movement decision perspective still remains unclear with most conclusions being drawn from visual interpretation or results of performance metrics. This work aims to study the EE trade-off of cognitive OSL actions by identifying movement patterns and quantify their values for multiple metrics associated with the decision. A large number of experiments were performed under realistic simulated environments, with the results showing the emergence of multiple well defined movement patterns and EE tendencies that remain identical across different scenarios.

ABSTRACT. Robotic assistive devices have emerged as a potential complement for repetitive and user-centered gait rehabilitation. In this field, the development of electromyography (EMG)-based torque controls has played a crucial role in improving the user experience with robotic assistive devices. However, most existing approaches for EMG-based joint torque estimation (i) are designed for upper limbs; (ii) often do not consider the complexity of the walking motion, focusing only on the stance phase; and (iii) rely on complex mathematical models that result in time-consuming estimations. This study aims to address these shortcomings by evaluating the generalization ability of a Deep Learning regressor (Convolutional Neural Network (CNN)) for estimating ankle torque trajectories, in real-time. Several inputs were incorporated, namely, EMG signals from Tibialis Anterior and Gastrocnemius Lateralis, hip kinematic data in the sagittal plane (angle, angular velocity, angular acceleration), walking speed (from 1.5 to 2.0 km/h), user's demographic (gender and age) and anthropometric information (height and mass, ranging from 1.50 to 1.90 m and 50.0 to 90.0 kg, respectively, and shank and foot lengths). Results showed that a CNN model with two convolutional layers showed the highest generalization ability (Root Mean Square Error: 23.4±8.36, Normalized Mean Square Error: 0.494±0.299, and Spearman Correlation 0.754±0.105). CNN model’s time-effectiveness was tested in an active ankle orthosis, being able to estimate ankle joint torques in less than 2 milliseconds. This study contributes to a more time-effective model for real-time EMG-based torque estimation, enabling a promising advancement in EMG-based torque control for lower limb robotic assistive devices.

ABSTRACT. This paper proposes a Potential Felds-based approach for coordinating a laparoscop-ic needle holder and grasper to achieve autonomous suturing in robotic laparoscopic surgery. This method allows seamless transition to teleoperated control based on Virtual Fixtures. The study presents three stages of suture, and experiments conduct-ed in a controlled surgical scenario test the feasibility of the proposed approach. The results indicate successful coordination between the needle holder and grasper, achieving a smooth and logical motion with minimal undesirable effects. The study also discusses challenges such as slow convergence towards attractive points and the need for accurate position determination, suggesting potential solutions and future research directions. Experiments demonstrate successful coordination between the tools, enabling smooth motion with minimal undesirable effects. The approach shows responsiveness and adaptability, promising enhanced autonomous robotic laparoscopy for various surgical tasks.

ABSTRACT. Laparoscopic surgery is a minimally invasive surgery that makes small incisions in the patient's tissue during the procedure in order to reduce infection and postoperative time. When using robots in this task applied to laparoscopic suturing, one of the main challenges encountered is human-robot iteration. To address this problem, this paper presents a functional architecture where the recognition of the surgeon's gestures by means of Machine Learning models and force-position control techniques of the surgical instrument stand out.

ABSTRACT. Dynamically avoiding collision is one of the key challenges to enabling industrial collaborative manufacturing. These techniques guarantee the safety of the operators without disregarding manufacturing efficiency. However, as these techniques usually rely on the jacobian, they can move the robot towards singular configurations, blocking its movements and affecting performance and application safety momentaneously. This work proposes a novel technique, \textit{wrist spherification}, to decouple the position and orientation of non-spherical wrist manipulators, enabling more efficient computation of online responses to dynamic obstacles. For such a purpose, the forward and inverse kinematics are proposed to address the position and velocity problems. The computational performance of the explicit inverse kinematics solutions proposed has been compared against the generic inverse kinematic solver of Matlab Robotics Toolbox based on Broyden-Fletcher-Goldfarb-Shanno (BFGS) gradient projection algorithm, reducing around five times the required mean computational time. A preliminary kinematic analysis is also presented, highlighting the kinematic improvement to compute offline singularities.