Abstract: A processor using a first Kalman filter estimates a host vehicle state from speed and yaw rate, the latter of which may be from a yaw rate sensor if speed is greater than a threshold, and, if less, from a steer angle sensor and speed. Road curvature parameters are estimated from a curve fit of a host vehicle trajectory or from a second Kalman filter for which a state variable may be responsive to a plurality of host state variables. Kalman filters may incorporate adaptive sliding windows. Curvature of a most likely road type is estimated with an interacting multiple model (IMM) algorithm using models of different road types. A road curvature fusion subsystem provides for fusing road curvature estimates from a plurality of curvature estimators using either host vehicle state, a map database responsive to vehicle location, or measurements of a target vehicle with a radar system.

Abstract: A first Kalman filter estimates true measures of yaw rate and vehicle speed from associated noisy measures thereof generated by respective sensors in a host vehicle, and a second Kalman filter estimates therefrom parameters of a clothoid model of road curvature. Measures of range, range rate, and azimuth angle from a target state estimation subsystem, e.g. a radar system, are processed by an extended Kalman filter to provide an unconstrained estimate of the state of a target vehicle. Associated road constrained target state estimates are generated for one or more roadway lanes, and are compared—either individually or in combination—with the unconstrained estimate. If a constrained target state estimate corresponds to the unconstrained estimate, then the state of the target vehicle is generated by fusing the unconstrained and constrained estimates; and otherwise is given by the unconstrained estimate alone.

Abstract: A plurality of events representative of a situation in which the host vehicle is operated are selected, including at least one set of related events. Input data is provided to an inference engine from either a first set of data representative of a target in a field of view of the host vehicle, a second set of data representative of the position or motion of the host vehicle, or a third set of data is representative of an environment of said host vehicle. The inference engine operates in accordance with an inference method to generate an output representative of a probability of occurrence of at least one event of the set of events, responsive to the input data, and possibly to one or more outputs at a past time. A countermeasure may be invoked responsive to one or more outputs from one or more inference engines.

Abstract: A radar system (14, 20, 22) generates a plurality of data points (x, y) representative of the position of a tracked object (48), and a representation of an associated path (50, 60) is formed therefrom. At least one quality measure, and at least one heading measure, of said representation, is calculated corresponding to at least one map coordinate (44). The quality and heading measures are stored in memory as a track map. Data from other tracked objects (58) is used to update the track map (38), resulting a plurality of heading values at associated map coordinates (44) that are representative of a path (42) followed by the tracked objects (48, 58).

Abstract: A radar system (14, 20, 22) generates a plurality of data points (x, y) representative of the position of a tracked object (48), and a representation of an associated path (50, 60) is formed therefrom. At least one quality measure, and at least one heading measure, of said representation, is calculated corresponding to at least one map coordinate (44). The quality and heading measures are stored in memory as a track map. Data from other tracked objects (58) is used to update the track map (38), resulting a plurality of heading values at associated map coordinates (44) that are representative of a path (42) followed by the tracked objects (48, 58).

Abstract: A neural network radar processor (10) comprises a multilayer perceptron neural network (100.1) comprising an input layer (102), a second layer (122), and at least a third layer (124), wherein each layer has a plurality of nodes (108), and respective subsets of nodes (108) of the second (122) and third (124) layers are interconnected so as to form mutually exclusive subnetworks (120). In-phase and quadrature phase time series from a sampled down-converted FMCW radar signal (19) are applied to the input layer, and the neural network (100) is trained so that the nodes of the output layer (106) are responsive to targets in corresponding range cells, and different subnetworks (120) are responsive to respectively different non-overlapping sets of target ranges. The neural network is trained with signals that are germane to an FMCW radar, including a wide range of target scenarios as well as leakage signals, DC bias signals, and background clutter signals.

Abstract: A plurality of events representative of a situation in which the host vehicle is operated are selected, including at least one set of related events. Input data is provided to an inference engine from either a first set of data representative of a target in a field of view of the host vehicle, a second set of data representative of the position or motion of the host vehicle, or a third set of data is representative of an environment of said host vehicle. The inference engine operates in accordance with an inference method to generate an output representative of a probability of occurrence of at least one event of the set of events, responsive to the input data, and possibly to one or more outputs at a past time. A countermeasure may be invoked responsive to one or more outputs from one or more inference engines.