This work relates to the automated parallel manipulation of parts at sub-millimeter scale and is a part of EU funded GOLEM Project. The main challenge at this scale is to develop novel methods for high throughput parallel assembly of components of a few hundreds of micrometers. At this scale, a serial approach would be extremely limited by the requirements on precision, speed ans especially by the particularities of physics. The proposed approach in this work is opto-fluidic, based on the Marangoni effect, a convective fluidic phenomena. The Marangoni effect is explored and analyzed both theoretically and experimentally. An experimental set-up is designed and constructed in this purpose. These studies show the advantages of the proposed approach for high speed manipulation of microcomponents in different sizes and geometries. The manipulation set-up is also entirely automated in order to show the parallel manipulation capabilities of this novel assembly technique. The first chapter gives an overview of contactless manipulation techniques at microscale, such as optical tweezers, electric field, dielectrophoresis, acoustic waves and thermal motion based techniques. A comparison of the techniques points Marangoni effect as a viable solution. The second chapter deals with the theoretical analysis of two convection phenomena: free convection and B'enard-Marangoni convection. This through a multi-physics finite elements based modeling. The governing equations for these phenomena are presented based on the fluid dynamics laws. A Proposed model is applied on a simple case of natural convection for initial analysis. Several simulations and their experimental validations are presented. Different parameters are analyzed such as water depth, temperature distribution and velocity field. Finally, a comparison between these phenomena is presented to know which mechanism predominates and is more suitable in our case. The Marangoni effect is presented as a promising method to drag micro-objects immersed in liquid media using only an IR laser beam as a heat source. This analysis allowed us to define the parameters for a conception of an experimental set-up for non-contact manipulation. The third chapter describes the design of this above mentioned robotic platform. This platform is composed of several components: an optical microscope, a laser source as local thermal source, a scanner to address the laser with precision and other electronics. A vision system, using a high speed camera is also implemented. A calibration of this vision system is established in order to define the available precision of the overall system, dimensions and measurable velocities of manipulated parts by experimental analysis. This approach also allows to measure instantaneous acceleration values and leads to the estimation of the force applied to manipulated objects. The fourth chapter deals with the automation of the manipulation process. The aim is to show that the proposed system is able to displace several microparts to predefined positions without user interaction. Particularly, the control of the Marangoni effect through the control of the position of the local heat source is demonstrated. The motion of this local thermal source is supplied by reflecting a laser beam on a mirror controlled by a high speed scanner. The implemented automation allows for a real time and high speed control hence it is possible to act simultaneously on several parts. The control loop is closed with vision feedback which is able to track at high frequency and sufficient precision all the involved parts at different form and dimensions. An experimental validation of parallel manipulation is describes and shows the originality of the proposed approach.