论文部分内容阅读
在文献中考虑利用多传感器跟踪机动目标一类的问题时,支持特定目标跟踪的传感器数量及类型通常相对于目标假定位置是固定的。然而,在许多多传感器系统中,支持某一特定目标跟踪的传感器数量及类型,可由于各个传感器的机动性、类型及资源的制约而随时变化。这种在传感器系统配置上的变化性,在跟踪机动目标时造成严重的问题,这是由于目标运动模型存在不确定性。卡尔曼滤波器通常用于滤波位置测量,以估计目标的位置,速度和加速度。在设计卡尔曼滤波器时,过程噪声(加速度)方差Qk的如此选定以致于65％到95％的概率区间能包含目标的最大加速度水平。然而,当目标机动时,加速度以一种确定性方式变化。于是,与过程噪声相关的白噪声假设发生偏离,滤波器在目标机动期间产生状态估计偏差。如果选定一个较大的Qk,则在机动时的状态估计偏差较小。但当目标不作机动时,此时的Qk只能粗劣地表征目标运动,而且滤波性能远远偏离最优了。这里,举出了目标在单一坐标系运动的例子,说明了利用多传感器跟踪机动目标存在的问题,从中表明两传感器(在确定条件下,其中包括各传感器的正确配置)具有较之单一传感器更糟糕的跟踪性能。将交互式多模型算法(IMM)应用于该范例中,证明了它是一种解决跟踪滤波器性能问题的潜在方法。
When considering the use of multiple sensors to track maneuvering targets in the literature, the number and type of sensors supporting a particular target tracking are often fixed relative to the target assumed position. However, in many multisensor systems, the number and type of sensors that support a particular target tracking can vary over time because of the mobility, type, and resource constraints of each sensor. This variability in the sensor system configuration poses a serious problem in tracking maneuver targets due to the uncertainty of the target motion model. Kalman filters are commonly used to filter position measurements to estimate the position, velocity, and acceleration of the target. In designing the Kalman filter, the process noise (acceleration) variance Qk is so selected that a probability interval of 65% to 95% can contain the target’s maximum acceleration level. However, when the target is maneuvering, the acceleration changes in a deterministic manner. The white noise assumption associated with the process noise then deviates, and the filter produces a state estimation bias during the target maneuver. If a larger Qk is selected, the state estimation deviation during maneuvering is smaller. However, when the target is not maneuvering, the Qk can only poorly characterize the target motion at this moment, and the filtering performance far deviates from the optimal one. Here, an example of moving a target in a single coordinate system is given, illustrating the problems with using multiple sensors to track maneuvering targets, showing that both sensors (under certain conditions, including the correct configuration of each sensor) have more advantages than single sensors Poor tracking performance. The use of an interactive multi-model algorithm (IMM) in this example proves that it is a potential solution to tracking performance issues in tracking filters.