DISA - LSI - Artículos en Congresos Internacionales

Permanent URI for this collection

http://dspace.uc3m.es/handle/10016/6476

Browse

Now showing 1 - 20 of 31

Guest editorial: Modeling and control of humanoid robots
(World Scientific Publishing, 2019-12) Monje Micharet, Concepción Alicia; Martínez de la Casa Díaz, Santiago
Evaluating the acceptance of autonomous vehicles in the future
(IEEE, 2023) Madridano Carrasco, Ángel; Gankhuyag, Delgermaa; Miguel Paraiso, Miguel Ángel de; Palos Lorite, Martin; Olaverri-Monreal, Cristina; Al-Kaff, Abdulla Hussein; Comunidad de Madrid; Ministerio de Ciencia e Innovación (España); Agencia Estatal de Investigación (España)
The continuous advance of the automotive industry is leading to the emergence of more advanced driver assistance systems that enable the automation of certain tasks and that are undoubtedly aimed at achieving vehicles in which the driving task can be completely delegated. All these advances will bring changes in the paradigm of the automotive market, as is the case of insurance. For this reason, CESVIMAP and the Universidad Carlos III de Madrid are working on an Autonomous Testing pLatform for insurAnce reSearch (ATLAS) to study this technology and obtain first-hand knowledge about the responsibilities of each of the agents involved in the development of the vehicles of the future. This work gathers part of the advancements made in ATLAS, which have made it possible to have an autonomous vehicle with which to perform tests in real environments and demonstrations bringing the vehicle closer to future users. As a result of this work, and in collaboration with the Johannes Kepler University Linz, the impact, degree of acceptance and confidence of users in autonomous vehicles has been studied once they have taken a trip on board a fully autonomous vehicle such as ATLAS. This study has found that, while most users would be willing to use an autonomous vehicle, the same users are concerned about the use of this type of technology. Thus, understanding the reasons for this concern can help define the future of autonomous cars.
Mono-DCNet: Monocular 3D Object Detection via Depth-based Centroid Refinement and Pose Estimation
(IEEE, 2022-07-19) Astudillo Olalla, Armando; Al Kaff, Abdulla Hussein Abdulrahman; García Fernández, Fernando
3D object detection is a well-known problem for autonomous systems. Most of the existing methods use sensor fusion techniques with Radar, LiDAR, and Cameras. However, one of the challenges is to estimate the 3D shape and location of the adjoining vehicles from a single monocular image without other 3D sensors; such as Radar or LiDAR. To solve the lack of the depth information, a novel method for 3D vehicle detection is presented. In this work, instead of using the whole depth map and the viewing angle (allocentric angle), only the depth mask of each object is used to refine the projected centroid and estimate its egocentric angle directly. The performance of the proposed method is tested and validated using the KITTI dataset, obtaining similar results to other state-of-the-art methods for Monocular 3D Object Detection.
Robot imitation through vision, kinesthetic and force features with online adaptation to changing environments
(IEEE, 2018-10-01) Fernández Fernández, Raúl; González Víctores, Juan Carlos; Estévez Fernández, David; Balaguer Bernaldo de Quirós, Carlos; Comunidad de Madrid
Continuous Goal-Directed Actions (CGDA)is a robot imitation framework that encodes actions as the changes they produce on the environment. While it presents numerous advantages with respect to other robot imitation frameworks in terms of generalization and portability, final robot joint trajectories for the execution of actions are not necessarily encoded within the model. This is studied as an optimization problem, and the solution is computed through evolutionary algorithms in simulated environments. Evolutionary algorithms require a large number of evaluations, which had made the use of these algorithms in real world applications very challenging. This paper presents online evolutionary strategies, as a change of paradigm within CGDA execution. Online evolutionary strategies shift and merge motor execution into the planning loop. A concrete online evolutionary strategy, Online Evolved Trajectories (OET), is presented. OET drastically reduces computational times between motor executions, and enables working in real world dynamic environments and/or with human collaboration. Its performance has been measured against Full Trajectory Evolution (FTE)and Incrementally Evolved Trajectories (IET), obtaining the best overall results. Experimental evaluations are performed on the TEO full-sized humanoid robot with “paint” and “iron” actions that together involve vision, kinesthetic and force features.
Reducing the number of evaluations required for CGDA execution through Particle Swarm Optimization methods
(IEEE, 2017-04-26) Fernández Fernández, Raúl; Estévez Fernández, David; González Víctores, Juan Carlos; Balaguer Bernaldo de Quirós, Carlos; Comunidad de Madrid
Continuous Goal Directed Actions (CGDA) is a robot learning framework that encodes actions as time series of object and environment scalar features. As the execution of actions is not encoded explicitly, robot joint trajectories are computed through Evolutionary Algorithms (EA), which require a large number of evaluations. The consequence is that evaluations are performed in a simulated environment, and the optimal robot trajectory computed is then transferred to the actual robot. This paper focuses on reducing the number of evaluations required for computing an optimal robot joint trajectory. Particle Swarm Optimization (PSO) methods have been adapted to the CGDA framework to be studied and compared: naíve PSO, Adaptive Fuzzy Fitness Granulation PSO (AFFG-PSO), and Fitness Inheritance PSO (FI-PSO). Experiments have been performed for two representative use cases within CGDA: the “wax” and the “painting” action. The experimental results of PSO methods are compared with those obtained with the Steady State Tournament used in the original proposal of CGDA. Conclusions extracted from these results depict a reduction of the number of required evaluations, with simultaneous tradeoff regarding the degree of fulfillment of the objective given by the optimization cost function.
Improving CGDA execution through genetic algorithms incorporating spatial and velocity constraints
(IEEE, 2017-04-26) Fernández Fernández, Raúl; Estévez Fernández, David; González Víctores, Juan Carlos; Balaguer Bernaldo de Quirós, Carlos; Comunidad de Madrid
In the Continuous Goal Directed Actions (CGDA) framework, actions are modelled as time series which contain the variations of object and environment features. As robot joint trajectories are not explicitly encoded in CGDA, Evolutionary Algorithms (EA) are used for the execution of these actions. These computations usually require a large number of evaluations. As a consequence of this, these evaluations are performed in a simulated environment, and the computed trajectory is then transferred to the physical robot. In this paper, constraints are introduced in the CGDA framework, as a way to reduce the number of evaluations needed by the system to converge to the optimal robot joint trajectory. Specifically, spatial and velocity constraints are introduced in the framework. Their effects in two different CGDA commonly studied use cases (the “wax” and “paint” actions) are analyzed and compared. The experimental results obtained using these constraints are compared with those obtained with the Steady State Tournament (SST) algorithm used in the original proposal of CGDA. Conclusions extracted from this study depict a high reduction in the required number of evaluations when incorporating spatial constraints. Velocity constraints provide however less promising results, which will be discussed within the context of previous CGDA works.
A method for synthetic LiDAR generation to create annotated datasets for autonomous vehicles perception
(IEEE, 2019-10-27) Beltrán de la Cita, Jorge; Cortes Lafuente, Irene; Barrera Del Pozo, Alejandro; Urdiales de la Parra, Jesús; Guindel Gómez, Carlos; García Fernández, Fernando; Escalera Hueso, Arturo de la; Comunidad de Madrid; Ministerio de Ciencia, Innovación y Universidades (España)
LiDAR devices have become a key sensor for autonomous vehicles perception due to their ability to capture reliable geometry information. Indeed, approaches processing LiDAR data have shown an impressive accuracy for 3D object detection tasks, outperforming methods solely based on image inputs. However, the wide diversity of on-board sensor configurations makes the deployment of published algorithms into real platforms a hard task, due to the scarcity of annotated datasets containing laser scans. We present a method to generate new point clouds datasets as captured by a real LiDAR device. The proposed pipeline makes use of multiple frames to perform an accurate 3D reconstruction of the scene in the spherical coordinates system that enables the simulation of the sweeps of a virtual LiDAR sensor, configurable both in location and inner specifications. The similarity between real data and the generated synthetic clouds is assessed through a set of experiments performed using KITTI Depth and Object Benchmarks.
Analysis of the Influence of Training Data on Road User Detection
(IEEE, 2018-09-12) Guindel Gómez, Carlos; Martín Gómez, David; Armingol Moreno, José María; Stiller, Christoph; Comunidad de Madrid
In this paper, we discuss the relevance of training data on modern object detectors used on onboard applications. Whereas modern deep learning techniques require large amounts of data, datasets with typical scenarios for autonomous vehicles are scarce and have a reduced number of samples. We conduct a comprehensive set of experiments to understand the effect of using a combination of two relatively small datasets to train an end-to-end object detector, based on the popular Faster R-CNN and enhanced with orientation estimation capabilities. We also test the adequacy of training models using partially available ground-truth labels, as a consequence of combining datasets aimed at different applications. Data augmentation is also introduced into the training pipeline. Results show a significant performance improvement in our exemplary case as a result of the higher variability of the training samples, thus opening a new way to improve the detection performance independently from the detector architecture.
Visual feature tracking based on PHD filter for vehicle detection
(IEEE, 2014-10-07) García, Fernando; Prioletti, Antonio; Cerri, Pietro; Broggi, Alberto; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Vehicle detection is one of the classical application among the Advance Driver Assistance Systems (ADAS). Applications like emergency braking or adaptive cruise control (ACC) require accurate and reliable vehicle detection. In latest years the improvements in vision detection have lead to the introduction of computer vision to detect vehicles by means of these more economical sensors, with high reliability. In the present paper, a novel algorithm for vehicle detection and tracking based on a probability hypothesis density (PHD) filter is presented. The first detection is based on a fast machine learning algorithm (Adaboost) and Haar-Like features. Later, the tracking is performed, by means features detected within the bounding box provided by the vehicle detection. The features, are tracked by a PHD filter. The results of the features being tracked are combined together in the last step, based on several different methods. Test provided show the performance of the PHD filter in public sequences using the different methods proposed.
Automatic laser and camera extrinsic calibration for data fusion using road plane
(IEEE, 2014-10-07) Rodríguez-Garavito, C. H.; Ponz Vila, Aurelio; García, Fernando; Martín Gómez, David; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Driving Assistance Systems and Autonomous Driving applications require trustable detections. These demanding requirements need sensor fusion to provide information reliable enough. But data fusion presents the problem of data alignment in both rotation and translation. Laser scanner and video cameras are widely used in sensor fusion. Laser provides operation in darkness, long range detection and accurate measurement but lacks the means for reliable classification due to the limited information provided. The camera provides classification thanks to the amount of data provided but lacks accuracy for measurements and is sensitive to illumination conditions. Data alignment processes require supervised and accurate measurements, that should be performed by experts, or require specific patterns or shapes. This paper presents an algorithm for inter-calibration between the two sensors of our system, requiring only a flat surface for pitch and roll calibration and an obstacle visible for both sensors for determining the yaw. The advantage of this system is that it does not need any particular shape to be located in front of the vehicle apart from a flat surface, which is usually the road. This way, calibration can be achieved at virtually any time without human intervention.
Continuous pose estimation for stereo vision based on UV disparity applied to visual odometry in urban environments
(IEEE, 2014) Musleh Lancis, Basam; Martín Gómez, David; Armingol Moreno, José María; Escalera Hueso, Arturo de la
Abstract: :This paper presents an autocalibration method to determine the pose of a stereo vision system based on knowing the geometry of the ground in front of the cameras. This pose changes considerably while the vehicle is driven, therefore it is good to know constantly the pose of the camera for several applications based on computer vision, such as advanced driver assistance systems, autonomous vehicles or robotics. These constant changes of the pose make interesting to be able to detect constantly the variations in its extrinsic parameters (height, pitch, roll). The validation of the autocalibration method is accomplished by a visual odometry implementation. A study of the improvement of the results of the visual odometry estimation taking into account the changes of the camera pose is presented, demonstrating the advantages of the autocalibration method.
SACAT: An instrumented vehicle for driver assistance and safety
(IEEE, 2012-09-05) Aliane, Nourdine; Fernández, Javier; Bemposta Rosende, Sergio; Mata Ortega, Mario; Diez Zaera, Ramiro
The present paper describes the framework and components of an instrumented vehicle for driver assistance and safety. The experimental platform is based on the use of an on-board computer vision system to capture the traffic signs, and on a multiple of electronic components to capture the vehicle state and identify drivers. The hardware architecture is designed with the purpose of making the deployment of functionalities related to driver assistance and road safety easy. The paper covers firstly the description of the hardware architecture, and then describes some of the implemented functionalities such as driver assistance based on traffic signs detection and recognition, traffic violation recorder, and a realization of an emergency call system.
Enhanced obstacle detection based on Data Fusion for ADAS applications
(IEEE, 2013) García, Fernando; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Abstract: Fusion is a common topic nowadays in Advanced Driver Assistance Systems (ADAS). The demanding requirements of safety applications require trustable sensing technologies. Fusion allows to provide trustable detections by combining different sensor devices, fulfilling the requirements of safety applications. High level fusion scheme is presented; able to improve classic ADAS systems by combining different sensing technologies i.e. laser scanner and computer vision. By means of powerful Data Fusion (DF) algorithms, the performance of classic ADAS detection systems is enhanced. Fusion is performed in a decentralized scheme (high level), allowing scalability; hence new sensing technologies can easily be added to increase the trustability and the accuracy of the overall system. Present work focus in the Data Fusion scheme used to combine the information of the sensors at high level. Although for completeness some details of the different detection algorithms (low level) of the different sensors is provided. The proposed work adapts a powerful Data Association technique for Multiple Targets Tracking (MTT): Joint Probabilistic Data Association (JPDA) to improve the trustability of the ADAS detection systems. The final application provides real time detection of road users (pedestrians and vehicles) in real road situations. The tests performed proved the improvement of the use of Data Fusion algorithms. Furthermore, comparison with other classic algorithms such as Global Nearest Neighbors (GNN) proved the performance of the overall architecture.
Part based pedestrian detection based on logic inference
(IEEE, 2013) Olmeda Reino, Daniel; Armingol Moreno, José María; Escalera Hueso, Arturo de la
This paper presents an approach on detection of largely occluded pedestrians. From a pair of synchronized cameras in the Visible Light (VL) and Far Infrared (FIR) spectrum individual detections are combined and final confidence is inferred using a small set of logic rules via a Markov Logic Network. Pedestrians not entirely contained in the image or occluded are detected based on the binary classification on subparts of the detection window. The presented method is applied to a pedestrian classification problem in urban environments. The classifier has been tested in an Intelligent Transportation System (ITS) platform as part of an Advanced Driver Assistance Systems (ADAS).
Context Aided Multilevel Pedestrian Detection
(IEEE, 2013) García, Fernando; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Abstract: The proposed work, depicts a novel algorithm able to provide multiple pedestrian detection, based on the use of classical sensors in modern automotive application and context information. The work takes advantage of the use of Joint Probabilities Data Association (JPDA) and context information to enhance the classic performance of the pedestrian detection algorithms. The combination of the different information sources with powerful tracking algorithms helps to overcome the difficulties of this processes, providing a trustable tool that improves performance of the single sensor detection algorithms. Context in a rich information source, able to improve the fusion process in all levels by the use of a priori knowledge of the application. In the present work multilevel fusion solution is provided for road safety application. Context is used in all the fusion levels, helping to improve the perception of the road environment and the relations among detections. By the fusion of all information sources, accurate and trustable detection is provided and complete situation assessment obtained, with estimation of the danger that involves any detection.
Joint Probabilistic Data Association fusion approach for pedestrian detection
(Ieee - The Institute Of Electrical And Electronics Engineers, Inc, 2013) García, Fernando; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Abstract: Fusion is becoming a classic topic in Intelligent Transport System (ITS) society. The lack of trustworthy sensors requires the combination of several devices to provide reliable detections. In this paper a novel approach, that takes advantage of the Joint Probabilistic Data Association technique (JPDA) for data association, is presented. The approach uses one of the most powerful techniques of Multiple Target Tracking theory and adapts it to fulfill the strong requirements of road safety applications. The different test performed proved that a powerful association technique can enhance the capacity of Advance Driver Assistance Systems. Two main sensors are used for pedestrian detection: laser scanner and computer vision. Furthermore, the approach takes advantage of the availability of other information sources i.e. context information and online information (GPS). The detections are fused using JPDA, enhancing the capacities of classical pedestrian detection systems, mainly based in visual information. The test performed also showed that JPDA improved the results offered by other data association techniques, e.g. Global Nearest Neighbors.
Data Fusion for Overtaking Vehicle Detection Based on Radar and Optical Flow
(IEEE, 2012-06-03) García, Fernando; Cerri, Pietro; Broggi, Alberto; Escalera Hueso, Arturo de la; Armingol Moreno, José María
Trustworthiness is a key point when dealing with vehicle safety applications. In this paper an approach to a real application is presented, able to fulfill the requirements of such demanding applications. Most of commercial sensors available nowadays are usually designed to detect front vehicles but lack the ability to detect overtaking vehicles. The work presented here combines the information provided by two sensors, a Stop&Go radar and a camera. Fusion is done by using the unprocessed information from the radar and computer vision based on optical flow. The basic capabilities of the commercial systems are upgraded giving the possibility to improve the front vehicles detection system, by detecting overtaking vehicles with a high positive rate.
Visual Ego Motion Estimation in Urban Environments based on U-V Disparity
(IEEE The Institute Of Electrical And Electronics Engineers, Inc, 2012-06-03) Musleh Lancis, Basam; Martín Gómez, David; Escalera Hueso, Arturo de la; Armingol Moreno, José María
The movement of the vehicle provides useful information for different applications, such as driver assistant systems or autonomous vehicles. This information can be known by means of a GPS, but there are some areas in urban environ ments where the signal is not available, as tunnels or streets with high buildings. A new method for 2D visual ego motion estimation in urban environments is presented in this paper. This method is based on a stereo-vision system where the feature road points are tracked frame to frame in order to estimate the movement of the vehicle, avoiding outliers from dynamic obstacles. The road profile is used to obtain the world coordinates of the feature points as a unique function of its left image coordinates. For these reasons it is only necessary to search feature points in the lower third of the left images. Moreover, the Kalman filter is used as a solution for filtering problem. That is, in some cases, it is necessary to filter raw data due to noise acquisition of time series. The results of the visual ego motion are compared with raw data from a GPS.
Specification and Development of a HMI for ADAS, Based in Usability and Accessibility Principles
(Ertico, 2010-10) Naranjo, J.E.; Jiménez, Felipe; García Fernández, Fernando; Armingol Moreno, José María; Zato, J.G.; Quero, A.
Traditionally, the design of road vehicle HMI is based in esthetic principles, maintaining it as an attractive factor for possible clients when buying a car. Only recently, ergonomic benefits have been applied to the design of HMIs, mainly following institutional impulses like the European Union one, but whose contribution is not clearly stated nowadays in commercial products. In this paper the authors present a study of the design of an HMI, based in usability and accessibility premises, centering the design in the user, as method to improve safety, making natural the communication with the driver as well as being able to transmitting information to the driver, from basic to the generated by ADAS installed in the car. Following these specifications a set of prototypes have been designed in order to develop a testbed that could be evaluated for a large set of drivers.
Novel method for vehicle and pedestrian detection based on information fusion
(Springer, 2013-03-11) García, Fernando; Escalera Hueso, Arturo de la; Armingol Moreno, José María
A novel approach for vehicle and pedestrian detection based on data fusion techniques is presented. The work fuses information from a 2D laser scanner and a computer camera, to provide detection and classification of vehicles and pedestrians in road environments. Thanks to the data fusion approach, the limitations of each sensor are overcome. Thus reliable system is provided, fulfilling the demands of road safety applications. Classification is performed using each sensor independently. Laser scanner approach is based in pattern matching and vision approach is based in the classical Histogram of Oriented Gradients features approach. A higher stage performs data fusion using Kalman Filter and Global Nearest Neighbors.

Browse

Recent Submissions