Steerable needles have been widely used in minimally-invasive surgical procedures such as brachytherapy, biopsy, and neurosurgery in which long flexible needles are inserted into the human body for treatment, diagnosis or sample removal. Prostate brachytherapy, known as a cost-effective and minimally invasive method, is a type of radiotherapy treatment which delivers radiation internally by implanting radioactive seeds at specific spots inside or around the prostate. Here accurate positioning of the needle plays a vital role as no proper treatment will happen if targeting accuracy is low. Moreover, poor placement of the radioactive seeds can cause long-term side effects due to the disruption of neighboring tissues or organs. Investigations reveal that using available techniques, even expert practitioners cannot achieve an average accuracy of less than 5 mm in seed placement. This may be improved by employing computer-controlled robots to compensate for the placement errors caused by needle deflection and tissue deformation.

This research aims to improve the needle placement accuracy using needle insertion robotic assistants. To this end, steering algorithms and controllers are employed in a feedback closed-loop structure. In these structures, the controllers are fed by needle tip deflection measurements, and perform any necessary control action to compensate the targeting errors. In this research different control strategies are proposed to compensate for the needle positioning errors. All the proposed controllers in this work are based on needle deflection measurements obtained from ultrasound images. To deal with system uncertainties, sliding mode control technique is used to control the needle tip position in 2D and 3D environments. Considering the planar case, a switching controller is proposed to compensate the needle tip deflection in one plane. Considering the trade-off between the performance (small targeting error) and cost (tissue trauma), a controller tuning method is provided which ensures the stability of the closed-loop system. This method is extended to the 3D environment using model averaging and input transformation technique to produce a multi-mode PWM representation of the system. Considering the constraints imposed by model averaging, proper controllers are designed and implemented to compensate for the tip positioning errors. To decrease the number of needle rotations and consequently decrease the tissue trauma, the switching controller is replaced by a sliding mode controller. This method considers two surfaces and finds the required rotation angle to compensate the tip defection error in 3D space. The proposed method is then improved by using an adaptive structure in which the sliding surfaces are rotated. Since ultrasound images can only provide measurements of the needle tip position and no information about the needle tip orientation can be retrieved from ultrasound images, model-based state observers are designed to overcome the measurement limitations and used in a feedback control scheme. The controllers are designed to meet the observer requirements and guarantee the stability of the system.As another step in compensating needle tip placement errors, we propose an extension to kinematic unicycle/bicycle model, which has been widely used for planning and control purposes. The goal of this modeling is to add new parameters to the existing model to account for tissue motion.