Bruno Marhic

Omnidirectional mirror design: multiple linear ring-windows viewing

In the following paper we present a methodology for designing catadioptric mirrors used as omnidirectional sensors taking info account linear projection properties. In order to get a constant resolution camera: such a linearity is stalled by pixel coordinates. This method also allows setting several areas to observe, and for each of them, the corresponding percentage of the mirrors radius. Then we illustrate our method with some simulation examples, and show matching real catadioptric mirrors. We include a software using this method. The former can be used to design and test the shape of a sensor, using the ray tracing software POV-Ray/sup trade;/.

Generic method for recognition of a wheelchair, even with a low resolution-effective sensor

In the following paper we present a method for recognition of special shapes. We aim at finding a wheelchair and its occupant in an omnidirectional image as fast as possible. We use two images to get a stereo vision and then compute the relative position of the object. Our method allows to detect it very quickly in a field of 360/spl deg/ in a single shot. Furthermore we can perform the detection without problem of orientation of the wheelchair towards the sensor.

Navigation paradigm of a prehensive robotics assistance

This article deals with a target tracking application for the disabled. The objective of this work is to track a wheelchair with a mobile platform and an embedded grasping arm (MANUS). We propose an approach based on an association of two Kalman filtering levels. The first level permits an estimation of the wheelchair configuration. The second is used to compute the mobile platform configuration in connection with its environment. The association of the two filtering process allows a robust tracking between two objects in movement.

Technical assistance: Two method for tracking of a wheelchair

This project deals with technical assistance for people of reduced mobility. We propose to study the technical of an original approach, which consists of mounting a prehensile arm onto a mobile base. The use of this mobile unit in relation to the patientpsilas wheelchair is carried out on a master-slave basis.. To ensure the tracking of the wheelchair by the mobile platform, we will present and compare two tracking approaches based on Kalman filterpsilas algorithm. The framework of the proposed algorithms for the tracking is named ldquoBearing only trackingrdquo or ldquoAngle only trackingrdquo. The first algorithm permits an estimation of the wheelchair configuration in its environment and the second uses a relative state configuration to compute the position and the velocity of the wheelchair with no prior knowledge of the environment. The proposed method allows a robust tracking between a mobile target (wheelchair) and a mobile observer (assistive platform).

Processing with omnidirectional vision sensors (some examples)

This paper deals with omnidirectional vision sensor. Through, the experience accumulates in our team, we want to evocate with the help of this conference, how omnidirectional vision sensors are useful in many applications. Such kinds of mirrors are not new and they were used for painting in the past. We will thus first propose a quick main history facts and then deal with some research project achieved in our group. The main purpose of this paper is not to focus on such or such processing but to give a general idea of what is it possible to handle with an omnidirectional vision system of perception

The mobile robot localization problem treated as a constraint satisfaction problem

Abstract The problem considered here is mobile robot localization. We emphasize particularly the localization imprecision estimation. To this aim, the imprecise measurements given by a telemetric sensor and two odometers mounted on the mobile robot are modelled by intervals and are fused by constraint propagation on intervals. Indeed, the localization problem from dead reckoning and exteroceptive data can be modelled as a constraint satisfaction problem (CSP). So, at the end of the localization process, we get a 3-D subpaving which is supposed to contain the robots position in a guaranteed way. The localization imprecision is naturally managed by our method.