Autonomous two-dimensional manipulation of microparticles to form patterns under an optical microscope by pushing and pulling them with a sharp nanoprobe is presented. Camera calibration is done by an iterative sliding-mode parameter observer designed to satisfy the Lyapunov stability criterion. Estimated parameters are also used to increase the bandwidth on position control. Individual particles are manipulated in linear trajectories, utilizing a procedure that allows the particles to stick near the tip for horizontal motion and to get released for vertical motion. Patterns (and assemblies) of particles are generated using a planning algorithm that orders individual manipulations based on the blockage of linear trajectories by other particles or target positions. Basically, the particle that blocks the most number of trajectories is positioned to the target position that blocks the least number of trajectories. This ordering algorithm is a minimization of blockages to the linear trajectories and hence, performs better than the conventional distance-based approaches. Efficiency of the approach is demonstrated on experimental results, positioning 4.5 mum diameter polystyrene particles on a flat glass substrate autonomously. Average accuracy of individual manipulations is less than 0.64 mum.