José María Cañas

Stereo Vision Tracking of Multiple Objects in Complex Indoor Environments

This paper presents a novel system capable of solving the problem of tracking multiple targets in a crowded, complex and dynamic indoor environment, like those typical of mobile robot applications. The proposed solution is based on a stereo vision set in the acquisition step and a probabilistic algorithm in the obstacles position estimation process. The system obtains 3D position and speed information related to each object in the robot’s environment; then it achieves a classification between building elements (ceiling, walls, columns and so on) and the rest of items in robot surroundings. All objects in robot surroundings, both dynamic and static, are considered to be obstacles but the structure of the environment itself. A combination of a Bayesian algorithm and a deterministic clustering process is used in order to obtain a multimodal representation of speed and position of detected obstacles. Performance of the final system has been tested against state of the art proposals; test results validate the authors’ proposal. The designed algorithms and procedures provide a solution to those applications where similar multimodal data structures are found.

Robotherapy with Dementia Patients

Humanoids have increasingly become the focus of attention in robotics research in recent years, especially in service and personal assistance robotics. This paper presents the application developed for humanoid robots in the therapy of dementia patients as a cognitive stimulation tool. The behaviour of the robot during the therapy sessions is visually programmed in a session script that allows music to play, physical movements (dancing, exercises, etc.), speech synthesis and interaction with the human monitor. The application includes the control software on board the robot and some tools like the visual script generator or a monitor to supervise the robot behaviour during the sessions. The robot applications impact on the patients health has been studied. Experiments with real patients have been performed in collaboration with a centre of research in neurodegenerative diseases. Initial results show a slight or mild improvement in neuropsychiatric symptoms over other traditional therapy methods.

Localization of legged robots combining a fuzzy-Markov method and a population of extended Kalman filters

This paper presents a new approach to robot vision-based self-localization in dynamic and noisy environments for legged robots when efficiency is a strong requirement. The major contribution of this paper is the improvement of a Markovian method based on a fuzzy occupancy grid (FMK). Our proposal combines FMK with a population of Extended Kalman Filters, making the complete algorithm both robust and accurate while keeping its computational cost bounded. Two different strategies have been designed to combine both the methods. They have been tested in the RoboCup environment and quantitatively compared with other approaches in several experiments with the real robot.

Dynamic Schema Hierarchies for an Autonomous Robot

This paper proposes a behavior based architecture for robot control which uses dynamic hierarchies of small schemas to generate autonomous behavior. Each schema is a flow of execution with a target, can be turned on and off, and has several parameters which tune its behavior. Low level schemas are woken up and modulated by upper level schemas, forming a hierarchy for a given behavior. At any time there are several awake schemas per level, running concurrently, but only one of them is activated by environment perception. When none or more than one schema wants to be activated then upper level schema is called for arbitration. This paper also describes an implementation of the architecture and its use on a real robot.

Visual fall detection for intelligent spaces

Artificial vision provides a remarkable good sensor when developing applications for intelligent spaces. Cameras are passive sensors that supply a great amount of information and are quite cheap. This paper presents an application for elderly care that detects falls or faints and automatically triggers the health alarm. It promotes the independent lifestyle of elder people at their homes as the monitoring application will call for timely health assistance when needed. The system extracts 3D information from several cameras and performs 3D tracking of the people in the intelligent space. It uses an evolutive multimodal algorithm to continuously estimate the 3D position in real time and learns the visual appearance of the persons. The system has been validated with some experiments in different real environments.

Perception and tracking of dynamic objects for optimization of avoidance strategies in autonomous piloting of vehicles

In the autonomous piloting of vehicles, the characterization of nearby dynamic object motion by perception and tracking techniques aids in the optimization of avoidance strategies. Knowledge of the features of object motion in goal-driven navigation allows for accurate deviation strategies to be implemented with appropriate anticipation. This perceptual competence is a fundamental issue in the design of unmanned commercial outdoor vehicles with an often reduced capability for maneuvering. To this aim, a grid map representation of the local panorama is proposed such that laser rangefinder images are converted into grid cells that are segmented and assigned to objects, allowing classification and monitoring. The motion properties of objects are thus used to establish avoidance behavior to smartly control the vehicle steering, such that a safe and optimal detour maneuver is carried out while driving to a target. The developed perceptual ability is demonstrated here in several tests performed in a relatively clutter-free area to detect and track walking pedestrians. Some results are also shown to highlight the modulation of moving object properties on trajectories described by a robot when a fuzzy avoidance strategy is used to control the steering angle.

An Overt Visual Attention Mechanism Based on Saliency Dynamics

Abstract A visual overt attention mechanism is presented in this paper. Our algorithm chooses the next fixation point for a robot mobile camera in order to track several objects around the robot simultaneously, even if the whole set of them can not be covered by the same camera image. Our approach is based on two related measurements, liveliness and saliency, that dynamically evolve depending on the image observations and the camera movements. The attention is shared among exploring the surroundings for salient features, reobserving the tracked objects and other task-dependent points to look at. A winner-takes-all competition provides a flexible time sharing behavior with natural appearances of new objects, disappearances and inhibition of return. It also accepts top down features to look for and different priorities for them. Several experiments have been carried out with a real Pioneer robot endowed with mobile camera and are also described.

From bio-inspired vs. psycho-inspired to etho-inspired robots

Artificial intelligence has been roughly divided into two schools of thought since its beginnings, the symbolic (also known as GOFAI, neat, classical, etc.) and the subsymbolic one (connectionist, scruffy, etc.). These two approaches have also had strong influence on the robotics field. In this paper we want to present a less known one, based on Ethology, and its application to the generation of autonomous behavior in mobile robots. In this way, we present the foundations of JDE (Jerarquia Dinamica de Esquemas), an etho-inspired architecture where behavior is organized as a dynamic hierarchy of independent schemas, as well as some examples of its applications, and a discussion about its properties.

Overt visual attention inside JDE control architecture

In this paper a visual overt attention mechanism is presented. It builds and keeps updated a scene representation of all the relevant objects around a robot, especially if they are far away of each other and do not lie in the same camera image. The algorithm chooses the next fixation point for a monocular camera, which is mounted over a pantilt unit. Our approach is based on two related dynamics: liveliness and saliency. The liveliness of each relevant object diminishes in time but increases with new observations of such object. The position of each valid object is a possible fixation point for the camera. The saliency of each fixation point increases in time but is reset after the camera visit such location. Real experiments with a pioneer robot endowed with a firewire camera on a pantilt unit are displayed

Comparison of Smart Visual Attention Mechanisms for Humanoid Robots

Cameras are one of the most relevant sensors in autonomous robots. One challenge with them is to manage the small field of view of regular cameras. A method of coping with this, similar to the attention systems in humans, is to use mobile cameras to cover all the robot surroundings and to perceive all the objects of interest to the robot tasks even if they do not lie in the same snapshot. A gaze control algorithm is then required that continuously selects where the camera should look. This paper presents three different covert attention mechanisms that have been designed and compared: one based on round-Robin sharing, another based on dynamic salience and one with fixed pattern camera movements. Several experiments have been performed with a humanoid robot in order to validate them and to give an objective comparison in the context of RoboCup, where the robots have several perceptive needs like localization and object tracking that must be satisfied and may not be fully compatible.