Antonio Marin Hernandez

Estimation of time-to-contact from Tau-margin and statistical analysis of behavior

Mobile robots need a measure for assessing the proximity of obstacles named Time-To-Contact. In this paper, we use the monocular vision in robot navigation in order to estimate the Time-To-Contact computing the Tau-margin by the ratio of change of apparent size of an obstacle. Our contribution is the proposal of a statistical method to predict the Time-To-Contact, where we collect data in different time slots. To give greater robustness in our estimations, it is proposed that the robot generates a linear approximation by removing the values that apparently do not share a linear behavior as the rest of the estimations. This is in order to external factors such as changes in lighting or problems when the robot is in motion, do not affect the estimation of the Time-To-Contact. Models and data obtained are compared and statistically analyzed visually and quantitatively. Experimental results from the application of the method on synthetic and real scenarios are reported.

An effective robotic model of action selection

In this paper we present a concise analysis of the requirements for effective action selection, and a centralized action selection model that fulfills most of these requirements. In this model, action selection occurs by combining sensory information from the non-homogenous sensors of an off-the-shelf robot with the feedback from competing behavioral modules. In order to successfully clean an arena, the animal robot (animat) has to present a coherent overall behavior pattern for both appropriate selection and termination of a selected behavior type. In the same way, an animat set in a chasing task has to present opportunist action selection to locate the nearest target. In consequence, both an appropriate switching of behavior patterns and a coherent overall behavior pattern are necessary for effective action selection.

Estimation of Time-to-Contact for Navigation of Autonomous Robots Using Parallel Processing

Given the trend of more complex tasks that robots should perform, we need to design increasingly robust solutions to perform tasks more efficiently and above all, real-time calculations. One of the most important tasks of a mobile robot is navigating through a room avoiding obstacles with which it may collide. Our contribution is to show how to compute efficiently time-to-contact for avoiding collisions in robotic navigation, using Graphics Processor Unit (GPU) and Compute Unified Device Architecture (CUDA). Description of the problem in different size of images and the complexity analysis of sequential and parallel ways are described. Finally, experimental results with synthetic and real images are shown.

Comparison and Analysis of Models to Predict the Motion of Segmented Regions by Optical Flow

Computer vision systems can predict the motion of objects if the movement behavior is analyzed over time, ie, it is possible to find out future values based on previously observed values. In this paper we present an statistical analysis which is aimed to compare two models to predict the position of moving objects in next frames. The models presented are the Kalman filter and an analysis of time series using an ARIMA model. Scenarios with different characteristics are presented as test cases. Segmentation of moving objects is done through the clustering of optical flow vectors for similarity, which are obtained by Pyramid Lucas and Kanade algorithm.