Research papers relying on Webots #2621
Replies: 39 comments
-
Neuroevolution of a Recurrent Neural Network for Spatial and Working Memory in a Simulated Robotic EnvironmentAnimals ranging from rats to humans can demonstrate cognitive map capabilities. We evolved weights in a biologically plausible recurrent neural network (RNN) using an evolutionary algorithm to replicate the behavior and neural activity observed in rats during a spatial and working memory task in a triple T-maze. The rat was simulated in the Webots robot simulator and used vision, distance and accelerometer sensors to navigate a virtual maze. After evolving weights from sensory inputs to the RNN, within the RNN, and from the RNN to the robot’s motors, the Webots agent successfully navigated the space to reach all four reward arms with minimal repeats before time-out. Our current findings suggest that it is the RNN dynamics that are key to performance, and that performance is not dependent on any one sensory type, which suggests that neurons in the RNN are performing mixed selectivity and conjunctive coding. Moreover, the RNN activity resembles spatial information and trajectory-dependent coding observed in the hippocampus. Collectively, the evolved RNN exhibits navigation skills, spatial memory, and working memory. Our method demonstrates how the dynamic activity in evolved RNNs can capture interesting and complex cognitive behavior and may be used to create RNN controllers for robotic applications. Hardware-and Situation-Aware Sensing for Robust Closed-Loop Control SystemsWhile vision is an attractive alternative to many sensors targeting closed-loop controllers, it comes with high timevarying workload and robustness issues when targeted to edge devices with limited energy, memory and computing resources. Replacing classical vision processing pipelines, e.g., lane detection using Sobel filter, with deep learning algorithms is a way to deal with the robustness issues while hardware-efficient implementation is crucial for their adaptation for safe closed-loop systems. However, while implemented on an embedded edge device, the performance of these algorithms highly depends on their mapping on the target hardware and situation encountered by the system. That is, first, the timing performance numbers (e.g., latency, throughput) depends on the algorithm schedule, i.e., what part of the AI workload runs where (e.g., GPU, CPU) and their invocation frequency (e.g., how frequently we run a classifier). Second, the perception performance (e.g., detection accuracy) is heavily influenced by the situation – e.g., snowy and sunny weather condition provides very different lane detection accuracy. These factors directly influence the closed-loop performance, for example, the lane-following accuracy in a lane-keep assist system (LKAS). We propose a hardware- and situation-aware design of AI perception where the idea is to define the situations by a set of relevant environmental factors (e.g., weather, road etc. in an LKAS).We design the learning algorithms and parameters, overall hardware mapping and its schedule taking the situation into account. We show the effectiveness of our approach considering a realistic LKAS case-study on heterogeneous NVIDIA AGX Xavier platform in a hardware-in-the-loop framework. Our approach provides robust LKAS designs with 32% better performance compared to traditional approaches. |
Beta Was this translation helpful? Give feedback.
-
Cooperative Architecture for Transportation System (CATS): Development of a Convoy Agent for (V /I)2C ApplicationsRecently, a significant advancement in Connected and Automated Vehicles (CAVs) is witnessed. Consequently, the development of the Vehicle-to-Everything (V2X) was arisen to establish the communication between the different agents in the system seeking some cooperative driving behavior. However, there is still a gap and various issues in handling the fleet of cars on highways due to the dynamic environment. Therefore, this paper proposes a new agent emergence in the transportation system that provides comprehensive cooperative behaviors. This agent is the Vehicular Convoy as an extension for the known platoons. However, this work is not only interested in studying the agent, but it also proposes a holistic Robotics Operating System (ROS)-based Cooperative Architecture for Transportation System known as (CATS). Three main agents -Car, Road, and Convoy- are combined in a unified framework, cooperating through the V2I/V2C/C2I paradigms. The architecture involves the different behaviors and necessary modules for each agent solely, in addition to developing the corresponding inter-communication protocols. The architecture is validated using Webots on different scenarios that demonstrate the capability of creating many convoys simultaneously, with different joining scenarios. The architecture performance shows promising results in achieving various cooperative behaviors successfully with practical computational complexity. |
Beta Was this translation helpful? Give feedback.
-
Linking global top-down views to first-person views in the brainHumans and other animals have a remarkable capacity to translate their position from one spatial frame of reference to another. The ability to seamlessly move between top-down and first-person views is important for navigation, memory formation, and other cognitive tasks. Evidence suggests that the medial temporal lobe and other cortical regions contribute to this function. To understand how a neural system might carry out these computations, we used variational autoencoders (VAEs) to reconstruct the first-person view from the top-down view of a robot simulation, and vice versa. Many latent variables in the VAEs had similar responses to those seen in neuron recordings, including location-specific activity, head direction tuning, and encoding of distance to local objects. Place-specific responses were prominent when reconstructing a first-person view from a top-down view, but head direction–specific responses were prominent when reconstructing a top-down view from a first-person view. In both cases, the model could recover from perturbations without retraining, but rather through remapping. These results could advance our understanding of how brain regions support viewpoint linkages and transformations. |
Beta Was this translation helpful? Give feedback.
-
Sim-to-Real Deep Reinforcement Learning for Safe End-to-End Planning of Aerial RobotsIn this study, a novel end-to-end path planning algorithm based on deep reinforcement learning is proposed for aerial robots deployed in dense environments. The learning agent finds an obstacle-free way around the provided rough, global path by only depending on the observations from a forward-facing depth camera. A novel deep reinforcement learning framework is proposed to train the end-to-end policy with the capability of safely avoiding obstacles. The Webots open-source robot simulator is utilized for training the policy, introducing highly randomized environmental configurations for better generalization. The training is performed without dynamics calculations through randomized position updates to minimize the amount of data processed. The trained policy is first comprehensively evaluated in simulations involving physical dynamics and software-in-the-loop flight control. The proposed method is proven to have a 38% and 50% higher success rate compared to both deep reinforcement learning-based and artificial potential field-based baselines, respectively. The generalization capability of the method is verified in simulation-to-real transfer without further training. Real-time experiments are conducted with several trials in two different scenarios, showing a 50% higher success rate of the proposed method compared to the deep reinforcement learning-based baseline. Keywords: deep reinforcement learning; obstacle avoidance; quadrotors; sim-to-real transfer. |
Beta Was this translation helpful? Give feedback.
-
DNN-based Accelerator for Intelligent Robotic Arm Control with High-Level SynthesisIntelligent robotics leverages deep learning to boost collaboration between humans and devices. Robotic controllers require a low-latency computation process for a real-time response when facing dynamic situations. Also, in the meantime, more controllers are designed with DNN-based reinforcement learning, which may need more computation power. In this paper, we use high-level synthesis to implement a DNN-based controller on an FPGA. The FPGA is built with an ESP SoC (System-on-Chip) platform, integrated with, and controlled through a host computer. We demonstrated the complete end-to-end controller system on a virtual robotic arm with 1041 times speedup compared with a CPU-based software implementation. |
Beta Was this translation helpful? Give feedback.
-
Indoor Stockpile Reconstruction Using Drone-Borne Actuated Single-Point LiDARsA low-cost, yet accurate approach for stockpile volume estimation within confined storage spaces is presented. The novel approach relies on actuating a single-point light detecting and ranging (1D LiDAR) sensor using a micro servo motor onboard a drone. The collected LiDAR ranges are converted to a point cloud that allows the reconstruction of 3D stockpiles, hence calculating the volume under the reconstructed surface. The proposed approach was assessed via simulations of a wide range of mission operating conditions while mapping two different stockpile shapes within the Webots robotic simulator. The influences from modulating the drone flight trajectory, servo motion waveform, flight speed, and yawing speed on the mapping performance were all investigated. For simple rectangular trajectories, it was found that having longer trajectories that are adjacent to the storage walls provides best reconstruction results with reasonable energy consumption. On the other hand, for short rectangular trajectories within the storage middle space, the yawing speed at corners must be decreased to ensure good reconstruction quality, although this can lead to relatively high energy consumption. Comparing the volumetric error values, the average error from the proposed 1D LiDAR system, when operating at 6°·s−1 maximum yawing speed at the corners, was 0.8 ± 1.1%, as opposed to 1.8 ± 1.7%, and 0.9 ± 1.0% from the 2D and 3D LiDAR options, respectively. Moreover, compared to 2D and 3D LiDARs, the proposed system requires less scanning speed for data acquisition, is much lighter, and allows a substantial reduction in cost. Keywords: drones, 1D LiDAR, stockpile volume estimation, confined space, aerial mapping. |
Beta Was this translation helpful? Give feedback.
-
Kick-motion Training with DQN in AI Soccer EnvironmentThis paper presents a technique to train a robot to perform kick-motion in AI soccer by using reinforcement learning (RL). In RL, an agent interacts with an environment and learns to choose an action in a state at each step. When training RL algorithms, a problem called the curse of dimensionality (COD) can occur if the dimension of the state is high and the number of training data is low. The COD often causes degraded performance of RL models. In the situation of the robot kicking the ball, as the ball approaches the robot, the robot chooses the action based on the information obtained from the soccer field. In order not to suffer COD, the training data, which are experiences in the case of RL, should be collected evenly from all areas of the soccer field over (theoretically infinite) time. In this paper, we attempt to use the relative coordinate system (RCS) as the state for training kick-motion of robot agent, instead of using the absolute coordinate system (ACS). Using the RCS eliminates the necessity for the agent to know all the (state) information of entire soccer field and reduces the dimension of the state that the agent needs to know to perform kick-motion, and consequently alleviates COD. The training based on the RCS is performed with the widely used Deep Q-network (DQN) and tested in the AI Soccer environment implemented with Webots simulation software. Keywords: Reinforcement learning (RL), Deep Q-Network (DQN), AI soccer, Curse of dimensionality (COD), Coordinate transformation matrix (CTM). |
Beta Was this translation helpful? Give feedback.
-
Human-Autonomous Teaming Framework Based on Trust ModellingWith the development of intelligent technology, autonomous agents are no longer just simple tools; they have gradually become our partners. This paper presents a trust-based human-autonomous teaming (HAT) framework to realize tactical coordination between human and autonomous agents. The proposed trust-based HAT framework consists of human and autonomous trust models, which leverage a fusion mechanism to fuse multiple performance metrics to generate trust values in real-time. To obtain adaptive trust models for a particular task, a reinforcement learning algorithm is used to learn the fusion weights of each performance metric from human and autonomous agents. The adaptive trust models enable the proposed trust-based HAT framework to coordinate actions or decisions of human and autonomous agents based on their trust values. We used a ball-collection task to demonstrate the coordination ability of the proposed framework. Our experimental results show that the proposed framework can improve work efficiency. Keywords: Trust model, Human-autonomous teaming, Reinforcement learning. |
Beta Was this translation helpful? Give feedback.
-
Balanced Standing on One Foot of Biped Robot Based on Three-Particle Model Predictive ControlBalancing is a fundamental task in the motion control of bipedal robots. Compared to two-foot balancing, one-foot balancing introduces new challenges, such as a smaller supporting polygon and control difficulty coming from the kinematic coupling between the center of mass (CoM) and the swinging leg. Although nonlinear model predictive control (NMPC) may solve this problem, it is not feasible to implement it on the actual robot because of its large amount of calculation. This paper proposes the three-particle model predictive control (TP-MPC) approach. It combines with the hierarchical whole-body control (WBC) to solve the one-leg balancing problem in real time. The bipedal robot’s torso and two legs are modeled as three separate particles without inertia. The TP-MPC generates feasible swing leg trajectories, followed by the WBC to adjust the robot’s center of mass. Since the three-particle model is linear, the TP-MPC requires less computational cost, which implies real-time execution on an actual robot. The proposed method is verified in simulation. Simulation results show that our method can resist much larger external disturbance than the WBC-only control scheme. Keywords: model predictive control, whole-body control, biped robot balance. |
Beta Was this translation helpful? Give feedback.
-
A Distributed Architecture for Onboard Tightly-Coupled Estimation and Predictive Control of Micro Aerial Vehicle FormationsThis paper presents a distributed estimation and control architecture for leader-follower formations of multi-rotor micro aerial vehicles. The architecture involves multi-rate extended Kalman filtering and nonlinear model predictive control in order to optimize the system performance while satisfying various physical constraints of the vehicles, such as actuation limits, safety thresholds and perceptual restrictions. The architecture leverages exclusively onboard sensing, computation, and communication resources and it has been designed for enhanced robustness to perturbations thanks to its tightly coupled components. The architecture has been initially tested and calibrated in a high-fidelity robotic simulator and then validated with a real two-vehicle system engaged in formation navigation and reconfiguration tasks. The results not only show the high formation performance of the architecture while satisfying numerous constraints but also indicate that it is possible to achieve full navigation and coordination autonomy in presence of severe resource constraints as those characterizing micro aerial vehicles. Keywords: Formation control, micro aerial vehicles, distributed non-linear model predictive control, relative and onboard localization, distributed estimation. |
Beta Was this translation helpful? Give feedback.
-
Linear and Nonlinear Model Predictive Control Strategies for Trajectory Tracking Micro Aerial Vehicles: A Comparative StudyThis paper presents a comparison of linear and nonlinear Model Predictive Control (MPC) strategies for trajectory tracking Micro Aerial Vehicles (MAVs). In this comparative study, we paid particular attention to establish quantitatively fair metrics and testing conditions for both strategies. In particular, we chose the most suitable numerical algorithms to bridge the gap between linear and nonlinear MPC, leveraged the very same underlying solver and estimation algorithm with identical parameters, and allow both strategies to operate with a similar computational budget. In order to obtain a well-tuned performance from the controllers, we employed the parameter identification results determined in a previous study for the same robotic platform and added a reliable disturbance observer to compensate for model uncertainties. We carried out a thorough experimental campaign involving multiple representative trajectories. Our approach included three different stages for tuning the algorithmic parameters, evaluating the predictive control feasibility, and validating the performances of both MPC-based strategies. As a result, we were able to propose a decisional recipe for selecting a linear or nonlinear MPC scheme that considers the predictive control feasibility for a peculiar trajectory, characterized by specific speed and acceleration requirements, as a function of the available on-board resources. |
Beta Was this translation helpful? Give feedback.
-
Constrained Visual Servoing of Quadrotors Based on Model Predictive ControlOne of the main issues of visual servoing schemes occurs when the target objects leave out the field of view (FOV) of the camera, which causes failure or poor performance of the controller. Solving this problem can be a challenge due to traditional controllers cannot include system's constraints. This work presents model predictive control (MPC) for constrained image-based visual servoing (IBVS) applied in quadrotors, considering constraints in FOV and restrictions in the control actions. To handle with image constraints the MPC considers: (1) the target objects stay only in camera's FOV and this work converts these restrictions in state constraints, (2) merge image instantaneous kinematics and the dynamic of commercial quadrotors (Mavic Pro 2) in a general mathematical model in order to satisfy the bounded control actions and image constraints. Due to commercial quadrotors allow velocities like control inputs, this work considers the reduced dynamic model in general velocities space and it was identified using Dynamic Mode Decomposition with control (DMDc) algorithm. This work uses Webots to evaluate the performance of the proposed controller. Finally the controller is compared with a classical IBVS scheme in order to verify the efficacy of the proposed controller and systematically evaluate the performance considering the system constraints. Keywords: Optimal Control, Non Linear Control Systems, Modelling, Identification, Signal Processing |
Beta Was this translation helpful? Give feedback.
-
Model Predictive Control for Full Autonomous Vehicle OvertakingDespite the many advancements in traffic safety, vehicle overtaking still poses significant challenges to both human drivers and autonomous vehicles, especially, how to evaluate the safety of passing a leading vehicle efficiently and reliably on a two-lane road. However, few realistic attempts in this field have been made in the literature to provide practical solutions without prior knowledge of the state of the environment and simplifications of vehicle models. These model simplifications make many of the proposed solutions in the literature unusable in real scenarios. Considering the dangers that can arise from performing a defective overtake and the substantial risk of vehicle crashes during high-speed maneuvers, in this paper we propose a system based on model predictive control to accurately estimate the safety of starting a vehicle overtake in addition to vehicle control during the maneuver that aims to ensure a collision-free overtake using a complete and realistic model of the vehicle’s dynamics. The system relies on a stereoscopic vision approach and machine learning techniques (YOLO and DeepSORT) to gather information about the environment such as lane width, lane center, and distance from neighboring vehicles. Furthermore, we propose a set of scenarios to test the performance of the proposed system based on accurate modeling of the environment under a range of traffic conditions and road architecture. The simulation result shows the high performance of the proposed system in ending collisions during overtaking and providing optimal pathing that minimizes travel time. |
Beta Was this translation helpful? Give feedback.
-
Deep Reinforcement Learning for Parameter Tuning of Robot Visual ServoingRobot visual servoing controls the motion of a robot through real-time visual observations. Kinematics is a key approach to Keywords: Embedded systems, Robotics, Robot visual servoing, kinematics, parameter tuning, deep reinforcement learning, knowledge transfer. |
Beta Was this translation helpful? Give feedback.
-
Development of a New Robust Stable Walking Algorithm for a Humanoid Robot Using Deep Reinforcement Learning with Multi-Sensor Data FusionThe difficult task of creating reliable mobility for humanoid robots has been studied for decades. Even though several different walking strategies have been put forth and walking performance has substantially increased, stability still needs to catch up to expectations. Applications for Reinforcement Learning (RL) techniques are constrained by low convergence and ineffective training. This paper develops a new robust and efficient framework based on the Robotis-OP2 humanoid robot combined with a typical trajectory-generating controller and Deep Reinforcement Learning (DRL) to overcome these limitations. This framework consists of optimizing the walking trajectory parameters and posture balancing system. Multi-sensors of the robot are used for parameter optimization. Walking parameters are optimized using the Dueling Double Deep Q Network (D3QN), one of the DRL algorithms, in the Webots simulator. The hip strategy is adopted for the posture Keywords: humanoid robot, stable walking, parameter optimization, deep reinforcement learning, multi-sensor. |
Beta Was this translation helpful? Give feedback.
-
Artificial intelligence in the Webots simulatorThe project explores the implementation of artificial intelligence in the Webots simulation environment by creating a basic mobile robot that can perform technological operations while in motion, with a focus on the development of soccer-playing robots. The team focused on achieving this goal using reinforcement learning and split into three subgroups to explore different approaches: classical reinforcement learning, pre-training, and the Deepbots framework. To ensure consistency and avoid potential inconsistencies in the reward system, the team created a common script that lists all the possible rewards an agent can earn based on the results of its actions. The results showed that all three groups successfully integrated with the Webots simulator and were able to learn from the simulation environment. The project provides insight into the potential effectiveness of different approaches for implementing artificial intelligence in the Webots simulator and demonstrates the ability to develop robots that can perform tasks in motion using machine learning-based techniques. Keywords: reinforcement learning, artificial intelligence, Webots, robot simulations, intelligent agent. |
Beta Was this translation helpful? Give feedback.
-
Poster: FedRos - Federated Reinforcement Learning for Networked Mobile-Robot CollaborationIn this paper, we propose FedRos, a Federated Reinforcement Learning based multi-robot system, which enables networked robots collaboratively to train a shared model without sharing their private sensing data. Firstly, we present the FedRos pipeline that embeds the Webots robotics simulator. We then highlight features of FedRos, including its compatibility with the state-of-the-art Federated Learning and Reinforcement Learning algorithms and its sim-to-real viability. Lastly, we present benchmark experiments to show the effectiveness of FedRos. Keywords: federated learning, robotics networks. |
Beta Was this translation helpful? Give feedback.
-
A Wheeled Mobile Robot Obstacles Avoidance for Navigation Control in a Static and Dynamic EnvironmentsThis research focuses on the performance of the obstacle avoidance feature implemented on a wheeled mobile robot by using Sugeno Fuzzy Inference System (FIS). The test was done using a simulation software Webots and the layout was first draft out using walls, floors, and objects in order to set a boundary and obstacles for the simulation. To test the effectiveness and efficiency in terms of time and distance, the simulation time and trajectory of the Khepera implementing the Sugeno FIS was recorded and compared with the Khepera without implementing the Sugeno FIS. The inputs of the ultrasonic sensors were also recorded in order to analyze whether if the Khepera detects incoming objects or if Khepera follows the rules set in the algorithm. |
Beta Was this translation helpful? Give feedback.
-
Transfer Learning for Embodied NeuroevolutionTransfer Learning (TL) has been widely used in machine learning where the neuronal layers in a learned source Artificial Neural Network (ANN) are transferred to a target ANN so as to speed up the latter’s learning. TL most often requires that the source and target domains are similar. However, its use in dissimilar domains as also in ANNs that use neuroevolution strategies has hardly been investigated. In this paper, we present a mechanism, suited for neuroevolution, that can identify specific neurons that need to be transferred. These Hot neurons from the source ANN, when transferred to the target ANN, helps in hastening the learning at the target. Simulations conducted using robots, clearly indicate that the mechanism is well suited for both similar and dissimilar tasks or environments. Keywords: artificial neural networks, embodied evolution, neuroevolution, robots, transfer learning. |
Beta Was this translation helpful? Give feedback.
-
Multimodal Interaction for Cobot Using MQTTFor greater efficiency, human–machine and human–robot interactions must be designed with the idea of multimodality in mind. To allow the use of several interaction modalities, such as the use of voice, touch, gaze tracking, on several different devices (computer, smartphone, tablets, etc.) and to integrate possible connected objects, it is necessary to have an effective and secure means of communication between the different parts of the system. This is even more important with the use of a collaborative robot (cobot) sharing the same space and very close to the human during their tasks. This study present research work in the field of multimodal interaction for a cobot using the MQTT protocol, in virtual (Webots) and real worlds (ESP microcontrollers, Arduino, IOT2040). We show how MQTT can be used efficiently, with a common publish/subscribe mechanism for several entities of the system, in order to interact with connected objects (like LEDs and conveyor belts), robotic arms (like the Ned Niryo), or mobile robots. We compare the use of MQTT with that of the Firebase Realtime Database used in several of our previous research works. We show how a “pick–wait–choose–and place” task can be carried out jointly by a cobot and a human, and what this implies in terms of communication and ergonomic rules, via health or industrial concerns (people with disabilities, and teleoperation). Keywords: multimodality, human–robot interaction, MQTT protocol, cobot, open-source, industrial IoT, virtual worlds, assistive technologies. |
Beta Was this translation helpful? Give feedback.
-
Scalability of cyber-physical systems in mixed reality experiences in ROS 2Nowadays, cyber-physical systems (CPSs) are composed of more and more agents and the demand for designers to develop ever larger multi-agent systems is a fact. When the number of agents increases, several challenges related to control or communication problems arise due to the lack of scalability of existing solutions. It is important to develop tools that allow control strategies evaluation of large-scale systems. In this paper, it is considered that a CPS is a heterogeneous robot multi-agent system that cooperatively performs a formation task through a wireless network. The goal of this research is to evaluate the system’s performance when the number of agents increases. To this end, two different mixed reality frameworks developed with the open-source tools Gazebo and Webots are used. These frameworks enable combining both real and virtual agents in a realistic scenario allowing scalability experiences. They also reduce the costs required when a significant number of robots operate in a real environment, as experiences can be conducted with a few real robots and a higher number of virtual robots by mimicking the real ones. Currently, the frameworks include several types of robots being the aerial robot Crazyflie 2.1 and differential mobile robots Khepera IV those used in this work. To illustrate the usage and performance of the frameworks, an event-based control strategy for rigid formations varying the number of agents is analyzed. The agents should achieve a formation defined by a set of desired Euclidean distances to their neighbors. To compare the scalability of the system in the two different tools, the following metrics have been used: formation error, CPU usage percentage, and the ratio between the real-time and the simulation time. The results show the feasibility of using Robot Operating System (ROS) 2 in distributed architectures for multi-agent systems in mixed reality experiences regardless of the number of agents and their nature. However, the two tools under study present different behaviors when the number of virtual agents grows in some of the parameters, and such discrepancies are analyzed. Keywords: Multi-Robot System, Mixed Reality, ROS 2, Formation Control |
Beta Was this translation helpful? Give feedback.
-
Towards enabling reliable immersive teleoperation through Digital Twin: A UAV command and control use caseThis paper addresses the challenging problem of enabling reliable immersive teleoperation in scenarios where an Unmanned Aerial Vehicle (UAV) is remotely controlled by an operator via a cellular network. Such scenarios can be quite critical particularly when the UAV lacks advanced equipment (e.g., Lidar-based auto stop) or when the network is subject to some performance constraints (e.g., delay). To tackle these challenges, we propose a novel architecture leveraging Digital Twin (DT) technology to create a virtual representation of the physical environment. This virtual environment accurately mirrors the physical world, accounting for 3D surroundings, weather constraints, and network limitations. To enhance teleoperation, the UAV in the virtual environment is equipped with advanced features that may be absent in the real UAV. Furthermore, the proposed architecture introduces an intelligent logic that utilizes information from both virtual and physical environments to approve, deny, or correct actions initiated by the UAV operator. This anticipatory approach helps to mitigate potential risks. Through a series of field trials, we demonstrate the effectiveness of the proposed architecture in significantly improving the reliability of UAV teleoperation. Keywords: Teleoperation, Digital Twin, 5G and beyond, Unmanned Aerial Vehicles, Internet of Things, Virtual Reality. |
Beta Was this translation helpful? Give feedback.
-
Learning Motion Skills for a Humanoid RobotThis thesis investigates the learning of motion skills for humanoid robots. As groundwork, a humanoid robot with integrated fall management was developed as an experimental platform. Then, two different approaches for creating motion skills were investigated. First, one that is based on Cartesian quintic splines with optimized parameters. Second, a reinforcement learning-based approach that utilizes the first approach as a reference motion to guide the learning. Both approaches were tested on the developed robot and on further simulated robots to show their generalization. A special focus was set on the locomotion skill, but a standing-up and kick skill are also discussed. |
Beta Was this translation helpful? Give feedback.
-
Modelling and Search-Based Testing of Robot Controllers Using Enzymatic Numerical P SystemsThe safety of the systems controlled by software is a very important area in a digitalized society, as the number of automated processes is increasing. In this paper, we present the results of testing the accuracy of different lane keeping controllers for an educational robot. In our approach, the robot is controlled using numerical P systems and enzymatic numerical P systems. For tests generation, we used an open-source tool implementing a search-based software testing approach. Keywords: tests generation, numerical P systems, enzymatic numerical P systems, search-based software testing, cyber-physical systems, membrane computing. |
Beta Was this translation helpful? Give feedback.
-
Deep reinforcement learning with action masking for differential-drive robot navigation using low-cost sensorsDriving a wheeled differential-drive robot to a target can be a complicated matter when trying to also avoid obstacles. Usually, such robots employ a variety of sensors, such as LiDAR, depth cameras, and others, that can be quite expensive. To this end, in this paper, we focus on a simple differential-drive wheeled robot that uses only inexpensive ultrasonic distance sensors and touch sensors. We propose a method for training a Reinforcement Learning (RL) agent to perform robot navigation to a target while avoiding obstacles. In order to increase the efficiency of the proposed approach we design appropriate action masks that can significantly increase the learning speed and effectiveness of the learned policy. As we experimentally demonstrated, the proposed agent can robustly navigate to a given target even in unknown procedurally generated environments, or even when denying part of its sensor input. Finally, we show a practical use-case using object detection to dynamically search for, and move to objects within unknown environments. The code used for conducted experiments is available online on Github. Keywords: robot navigation, low-cost robot sensors, deep reinforcement learning, action masking. |
Beta Was this translation helpful? Give feedback.
-
Optimization and Evaluation of Multi Robot Surface Inspection Through Particle Swarm OptimizationRobot swarms can be tasked with a variety of automated sensing and inspection applications in aerial, aquatic, and surface environments. In this paper, we study a simplified two-outcome surface inspection task. We task a group of robots to inspect and collectively classify a 2D surface section based on a binary pattern projected on the surface. We use a decentralized Bayesian decision-making algorithm and deploy a swarm of miniature 3-cm sized wheeled robots to inspect randomized black and white tiles of 1m × 1m. We first describe the model parameters that characterize our simulated environment, the robot swarm, and the inspection algorithm. We then employ a noise-resistant heuristic optimization scheme based on the Particle Swarm Optimization (PSO) using a fitness evaluation that combines decision accuracy and decision time. We use our fitness measure definition to asses the optimized parameters through 100 randomized simulations that vary surface pattern and initial robot poses. The optimized algorithm parameters show up to a 55% improvement in median of fitness evaluations against an empirically chosen parameter set. |
Beta Was this translation helpful? Give feedback.
-
Initial Task Assignment in Multi-Human Multi-Robot Teams: An Attention-enhanced Hierarchical Reinforcement Learning ApproachMulti-human multi-robot teams (MH-MR) obtain tremendous potential in tackling intricate and massive missions by merging distinct strengths and expertise of individual members. The inherent heterogeneity of these teams necessitates advanced initial task assignment (ITA) methods that align tasks with the intrinsic capabilities of team members from the outset. While existing reinforcement learning approaches show encouraging results, they might fall short in addressing the nuances of long-horizon ITA problems, particularly in settings with large scale MH-MR teams or multifaceted tasks. To bridge this gap, we propose an attention-enhanced hierarchical reinforcement learning approach that decomposes the complex ITA problem into structured sub-problems, facilitating more efficient allocations. To bolster sub-policy learning, we introduce a hierarchical cross-attribute attention (HCA) mechanism, encouraging each sub-policy within the hierarchy to discern and leverage the specific nuances in the state space that are crucial for its respective decision-making phase. Through an extensive environmental surveillance case study, we demonstrate the benefits of our model and the HCA inside. Experimental details are available |
Beta Was this translation helpful? Give feedback.
-
Deep Active Perception for Object Detection using Navigation ProposalsDeep Learning (DL) has brought significant advances to robotics vision tasks. However, most existing DL methods have a major shortcoming - they rely on a static inference paradigm inherent in traditional computer vision pipelines. On the other hand, recent studies have found that active perception improves the perception abilities of various models by going beyond these static paradigms. Despite the significant potential of active perception, it poses several challenges, primarily involving significant changes in training pipelines for deep learning models. To overcome these limitations, in this work, we propose a generic supervised active perception pipeline for object detection that can be trained using existing off-the-shelf object detectors, while also leveraging advances in simulation environments. To this end, the proposed method employs an additional neural network architecture that estimates better viewpoints in cases where the object detector confidence is insufficient. The proposed method was evaluated on synthetic datasets - constructed within the Webots robotics simulator -, showcasing its effectiveness in two object detection cases. Keywords: active object detection, active perception, deep learning. |
Beta Was this translation helpful? Give feedback.
-
Visual Servoing NMPC Applied to UAVs for Photovoltaic Array InspectionThe photovoltaic (PV) industry is seeing a significant shift toward large-scale solar plants, where traditional inspection methods have proven to be time-consuming and costly. Currently, the predominant approach to PV inspection using unmanned aerial vehicles (UAVs) is based on photogrammetry. However, the photogrammetry approach presents limitations, such as an increased amount of useless data during flights, potential issues related to image resolution, and the detection process during high-altitude flights. In this work, we develop a visual servoing control system applied to a UAV with dynamic compensation using a nonlinear model predictive control (NMPC) capable of accurately tracking the middle of the underlying PV array at different frontal velocities and height constraints, ensuring the acquisition of detailed images during low-altitude flights. The visual servoing controller is based on the extraction of features using RGB-D images and the Kalman filter to estimate the edges of the PV arrays. Furthermore, this work demonstrates the proposal in both simulated and real-world environments using the commercial aerial vehicle (DJI Matrice 100), with the purpose of showcasing the results of the architecture. Our approach is available for the scientific community in: https://github.com/EPVelasco/VisualServoing_NMPC |
Beta Was this translation helpful? Give feedback.
-
Teaching practical robotics during the COVID-19 pandemic: a case study on regular and hardware-in-the-loop simulationsLaboratory experiments are important pedagogical tools in engineering courses. Restrictions related to the COVID-19 pandemic made it very difficult or impossible for laboratory classes to take place, resulting on a fast transition to simulation as an approach to guarantee the effectiveness of teaching. Simulation environments are powerful tools that can be adopted for remote classes and self-study. With these Keywords: hardware-in-the-loop simulation, robot simulation, engineering education, distance learning, COVID-19 pandemic. |
Beta Was this translation helpful? Give feedback.
-
There was recently quite a few scientific publications based on Webots which may be of interest for the community:
A Fuzzy Cooperative Localisation Framework for Underwater Robotic Swarms
This article proposes a holistic localisation framework for underwater robotic swarms to dynamically fuse multiple position estimates of an autonomous underwater vehicle while using fuzzy decision support system. A number of underwater localisation methods have been proposed in the literature for wireless sensor networks. The proposed navigation framework harnesses the established localisation methods in order to provide navigation aids in the absence of acoustic exteroceptive sensors navigation aid (i.e., ultra-short base line) and it can be extended to accommodate newly developed localisation methods by expanding the fuzzy rule base. Simplicity, flexibility, and scalability are the main three advantages that are inherent in the proposed localisation framework when compared to other traditional and commonly adopted underwater localisation methods, such as the Extended Kalman Filter. A physics-based simulation platform that considers environment’s hydrodynamics, industrial grade inertial measurement unit, and underwater acoustic communications characteristics is implemented in order to validate the proposed localisation framework on a swarm size of 150 autonomous underwater vehicles. The proposed fuzzy-based localisation algorithm improves the entire swarm mean localisation error and standard deviation by 16.53% and 35.17%, respectively, when compared to the Extended Kalman Filter based localisation with round-robin scheduling.
Evolutionary Gait Transfer of Multi-Legged Robots in Complex Terrains
Robot gait optimization is the task of generating an optimal control trajectory under various internal and external constraints. Given the high dimensions of control space, this problem is particularly challenging for multi-legged robots walking in complex and unknown environments. Existing literature often regard the gait generation as an optimization problem and solve the gait optimization from scratch for robots walking in a specific environment. However, such approaches do not consider the use of pre-acquired knowledge which can be useful in improving the quality and speed of motion generation in complex environments. To address the issue, this paper proposes a transfer learning-based evolutionary framework for multi-objective gait optimization, named Tr-GO. The idea is to initialize a high quality population by using the technique of transfer learning, so any kind of population-based optimization algorithms can be seamlessly integrated into this framework. The advantage is that the generated gait can not only dynamically adapt to different environments and tasks, but also simultaneously satisfy multiple design specifications (e.g., speed, stability). The experimental results show the effectiveness of the proposed framework for the gait optimization problem based on three multi-objective evolutionary algorithms: NSGA-II, RM-MEDA and MOPSO. When transferring the pre-acquired knowledge from the plain terrain to various inclined and rugged ones, the proposed Tr-GO framework accelerates the evolution process by a minimum of 3-4 times compared with non-transferred scenarios.
High-Speed Autonomous Robotic Assembly Using In-Hand Manipulation and Re-Grasping
This paper presents an autonomous robotic assembly system for Soma cube blocks, which, after observing the individual blocks and their assembled shape, quickly plans and executes the assembly motion sequence that picks up each block and incrementally build the target shape. A multi stage planner is used to find the suitable assembly solutions, assembly sequences and grip sequences considering various constraints, and re-grasping is used when the block target pose is not directly realizable or the block pose is ambiguous. The suggested system is implemented for a commercial UR5e robotic arm and a novel two degrees of freedom (DOF) gripper capable of in-hand manipulation, which further speeds up the manipulation speed. It was experimentally validated through a public competitive demonstration, where the suggested system completed all assembly tasks reliably with outstanding performance.
Feel free to post here more Webots-related publications.
Beta Was this translation helpful? Give feedback.
All reactions