Robotic control of needle bending aims at increasing the precision of percutaneous procedures. Ultrasound feedback is preferable for its clinical ease of use, cost and compactness but raises needle detection issues. In this paper, we propose a complete system dedicated to robotized guidance of a flexible needle under 3D ultrasound imaging. This system includes a medical robot dedicated to transperineal needle positioning and insertion, a rapid path planning for needle steering using bevel-tip needle natural curvature in tissue, and an ultrasound-based automatic needle detection algorithm. Since ultrasound-based automatic needle steering is often made difficult by the needle localization in biological tissue, we quantify the benefit of using flexible echogenic needles for robotized guidance under 3D ultrasound. The ``echogenic'' term refers to the etching of microstructures on the needle shaft. We prove that these structures improve needle visibility and detection robustness in ultrasound images. We finally present promising results when reaching targets using needle steering. The experiments were conducted with various needles in different media (synthetic phantoms and ex-vivo biological tissue). For instance, with nitinol needles the mean accuracy is 1.2mm (respectively 3.8mm) in phantoms (resp. biological tissue).