The detection of a pattern in an audio sequence is considered. An approach relying on the Stochastic Matched Filtering theory is proposed. It consists in first defining offline a basis from the statistics of the pattern and of the noise, then in isolating the pattern by means of a likelihood ratio test involving the online decomposition of the audio sequence on this basis. A simulated case study is proposed, which provides some guidelines to the tuning of the algorithm. Then, experimental results concerning the application of the method to voice activity detection are presented.

We propose a systematic modelling of the nonholonomic mobile manipulators built from a robotic arm mounted on a wheeled mobile platform. It extends the fundamental notions of nonholonomy, mobility and maneuverability to the case of these hybrid holonomic/nonholonomic systems. It offers unambiguous definitions and models to the designer of kinematic control laws.

We propose a generic scheme to solve the kinematic control problem of wheeled mobile manipulators when the operational motion is imposed. We generalize the Additional Task Method to solve the control problem of these redundant nonholonomic systems.

We extend the standard definition of manipulability to the case of a nonholonomic mobile manipulator built from an n joint robotic arm and a nonholonomic mobile platform. The effects of mounting the arm on a nonholonomic platform are shown through the analysis of the manipulability thus defined. Applications of criteria inherited from manipulability considerations are given to justify design and to generate the controls of our system.

Deals with path planning for mobile nonholonomic manipulators. The system is composed of a nonholonomic mobile platform and a simple kinematic chain holonomic arm. We propose strategies for planning paths between two points in the generalized and operational spaces. Redundancy and optimization criteria are key elements of the approach. Applications follow for a planar system in the case of point to point tasks.

This paper deals with the optimal control (time optimal) of a mobile robot moving in an unobstructed planar environment. The mobile robot is equipped with two independent acceleration-driven wheels so that accelerations are bounded. Using the mobile robot's configuration known through its position and orientation, the problem consists of determining the optimal control that allows the robot to move between two imposed configurations, the start and arrival taking place at zero velocity. Our approach is based on analytically extracting as much information as possible from the Pontryagin's maximum principle (PMP) before starting a numerical search. The recently reported numerical result, showing that this extremal control which is of "bang-bang" type and only requires four switch times or five sequences, is invalidated. We show a configuration space zone in which no extremal control using five sequences is optimal. We also show that there exist, in a part of this zone, extremal controls requiring seven sequences that are candidates for optimality.

Real-time dynamic control of robot manipulators requires on-line computation of the dynamic model that expresses the generalized forces to be applied to their joints, as a function of their generalized coordinates, velocities and accelerations. To do this, this paper presents a method of computation of this model that uses a quasiminimal number of elementary arithmetical operations and that can be applied systematically to robot manipulators with a simple kinematic chain structure and revolute and/or prismatic joints. To reach this quasi-minimal number, use if primarily made of the following : * a computation that is intrinsic rather than extrinsic, analytical rather than numerical and iterative rather than developed ; * the Newton-Euler formalism rather than the Lagrangian one and * the notion of augmented body, generalized to this type of structure. An example demonstrates that the computation of the dynamic model of an industrial robot manipulator with six revolute joints (the most complicated case in practice) can be effected with less than 300 arithmetical operations (adds and multiplies).