Abstract: To address the low tracking performance of the traditional probability hypothesis density/cardinalized probability hypothesis density (PHD/CPHD) filter in complex scenarios, such as unknown dynamic parameters (e.g., maneuver parameters and process noise) and unknown newborn targets, we propose an adaptive grid-driven (GD) PHD/CPHD filter algorithm. First, we divide the tracking area into an initial grid. Then, we identify newborn targets based on the measurement information, update the weights of the grids, and then construct grid regions based on these weights to extract the target states. Finally, we propose a method for expanding the grid area based on the target speed by adaptively re-dividing the expanded grid area based on the grid resolution to achieve multitarget tracking in complex scenarios. The experimental results show that when the newborn and surviving targets are within the ideal distance, the proposed algorithm performs better than the traditional particle-filter-based PHD method in terms of the average optimal sub-pattern assignment distance and the estimation of the average target number for tracking multitargets with unknown dynamic parameters and unknown newborn targets.