Reliable and efficient perception in dynamic environments especially in densely cluttered environments is still a challenge today. Most of the systems used for target tracking are based on object models. Such approaches usually fail in complex environments with a variety of different moving obstacles. In this report, we propose a new non object based model to perform tracking in such environments. The approach is called Bayesian Occupancy Filtering (BOF). It basically consists of regular grids where each cell contains information on the distribution on the grid occupancy and velocity. The cells are assumed to be independent from one another to avoid combinatorial explosion. The estimation of cell occupancy and cell velocity are performed in a manner similar to classical filtering, where there is a prediction and estimation stage. Sensory data from different sensors can be fused onto BOF cells and since cells are independent from one another, the notion of physical objects does not exist in this space. Doing so avoids the problem of data association which have to be often resolved in object based tracking systems. This report describes how the filtering is performed on this grid space in the context of target tracking. The disadvantage of tracking in grid space is when the cells move with a velocity which results in a cell position that does not end in only one single cell. A proposed method of handling such discretisation by staggering the updates of the cell such that cells are updated only when the fit is perfect. This corresponds to the intuitive notion of updating cells with higher velocity at a higher frequency. Such attention focusing enables the reduction of computation burden.