Abstract: In an embodiment, a method of training a detection transformer (DETR) comprises: initializing, with at least one processor, parameters of the DETR; propagating, with the at least one processor, an input image through the DETR; determining, with the at least one processor, error values by comparing the output of the DETR with known expected output; and iteratively updating, with the at least one processor, the parameters in the DETR based on the error values by minimizing a global loss, wherein minimizing the global loss comprises generating an assignment cost matrix; solving an assignment optimization problem based on the assignment cost matrix to obtain an optimal assignment cost, and then minimizing a global loss function that is expressed as a sum of the optimal assignment cost and an assignment-independent loss.

Abstract: The present disclosure includes systems, methods, and computer-readable storage media facilitating multi-target tracking with dependent likelihood structures. To facilitate tracking, sensor data associated with an environment that a vehicle is located in may be received. A set of candidate hypotheses may be obtained based on the sensor data. Each candidate hypothesis in the set of candidate hypotheses may include track and measurement information representative of candidate locations for one or more targets. The set of candidate hypotheses are evaluated against at least one criterion. A ranked set of hypotheses are determined based on the evaluation of the set of candidate hypothesis against the at least one criterion. The ranked set of hypotheses may include a subset of hypotheses selected from among the set of candidate hypotheses and may be used to estimate a location of each of the one or more targets.

Abstract: A technique for tracking objects includes: determining a set of detected measurements based on a received return signal; determining a group that includes a set of group measurements and a set of group tracks; creating a merged factor, including a merged set of track state hypotheses associated with a merged set of existing tracks including a first set of existing tracks and a second set of existing tracks, by calculating the cross-product of a first set of previous track state hypotheses and a second set of previous track state hypotheses; determining a first new factor and a second new factor; calculating a first set of new track state hypotheses for the first new factor based on a first subset of the group measurements; and calculating a second set of new track state hypotheses for the second new factor based on a second subset of the group measurements.

Abstract: A technique for tracking objects includes: determining a set of detected measurements based on a received return signal; determining a group that includes a set of group measurements and a set of group tracks; creating a merged factor, including a merged set of track state hypotheses associated with a merged set of existing tracks including a first set of existing tracks and a second set of existing tracks, by calculating the cross-product of a first set of previous track state hypotheses and a second set of previous track state hypotheses; determining a first new factor and a second new factor; calculating a first set of new track state hypotheses for the first new factor based on a first subset of the group measurements; and calculating a second set of new track state hypotheses for the second new factor based on a second subset of the group measurements.

Abstract: Provided are methods for object tracking, which can include receiving sensor data characterizing respective detected objects. The methods can also include generating a data structure based on the data characterizing the respective detected objects. The data structure can include a graph of nodes representing states of the objects and edges representing hypothetical transitions in states of the objects. The methods can also include applying a predictive model to the data structure. The predictive model can be trained to receive the state as inputs and produce an identification of a set of nodes and edges corresponding to the one of the respective detected objects. The methods can further include providing data based on the identification of the set of nodes and edges to a planning system of the vehicle and causing the vehicle to operate based on providing the data. Systems and computer program products are also provided.

Abstract: A technique for tracking objects includes: determining a set of detected measurements based on a received return signal; determining a group that includes a set of group measurements and a set of group tracks; creating a merged factor, including a merged set of track state hypotheses associated with a merged set of existing tracks including a first set of existing tracks and a second set of existing tracks, by calculating the cross-product of a first set of previous track state hypotheses and a second set of previous track state hypotheses; determining a first new factor and a second new factor; calculating a first set of new track state hypotheses for the first new factor based on a first subset of the group measurements; and calculating a second set of new track state hypotheses for the second new factor based on a second subset of the group measurements.

Abstract: The present disclosure includes systems, methods, and computer-readable storage media facilitating multi-target tracking with dependent likelihood structures. To facilitate tracking, sensor data associated with an environment that a vehicle is located in may be received. A set of candidate hypotheses may be obtained based on the sensor data. Each candidate hypothesis in the set of candidate hypotheses may include track and measurement information representative of candidate locations for one or more targets. The set of candidate hypotheses are evaluated against at least one criterion. A ranked set of hypotheses are determined based on the evaluation of the set of candidate hypothesis against the at least one criterion. The ranked set of hypotheses may include a subset of hypotheses selected from among the set of candidate hypotheses and may be used to estimate a location of each of the one or more targets.

Abstract: A technique for tracking objects includes: determining a set of detected measurements based on a received return signal; determining a group that includes a set of group measurements and a set of group tracks; creating a merged factor, including a merged set of track state hypotheses associated with a merged set of existing tracks including a first set of existing tracks and a second set of existing tracks, by calculating the cross-product of a first set of previous track state hypotheses and a second set of previous track state hypotheses; determining a first new factor and a second new factor; calculating a first set of new track state hypotheses for the first new factor based on a first subset of the group measurements; and calculating a second set of new track state hypotheses for the second new factor based on a second subset of the group measurements.

Abstract: A receiver includes: a filter circuit to generate predicted measurements for a set of tracks based on previous kinematic states of the tracks and timing and source data of next compressed measurements associated to the tracks, generate probability data of differences between the predicted measurements and next measurement data, generate the next measurement data using the predicted measurements and quantized differences between the predicted measurements and the next measurement data, and generate next kinematic states of the tracks based on the previous kinematic states, the timing and source data of the next compressed measurements, and the generated measurement data; a quantizer circuit to quantize the probability data into quantization tables and look up the quantized differences from corresponding indices in the quantization tables; and a decoder circuit to decode encoded index data of the next compressed measurements into the corresponding indices in the quantization tables using the quantized probabili

Abstract: Solutions for object tracking problems are presented by gathering images using one or more cameras, processing the gathered images to generate a directed acyclic graph, using the directed acyclic graph to determine a path cover that achieves maximum weight and satisfies one or more positive or negative constraints, and using the path cover to solve the object tracking problem. A first set of solutions utilizes trellis graphs, a second set of solutions employs a greedy approach, and a third set of solutions uses search algorithms.

Abstract: A system and method for projecting target tracks produced by a remote tracker onto a surface of the Earth to obtain projected tracks is provided. Projection bias in a projected track from the remote tracker projected onto a planar map is removed by computing a discrete Frechet distance from a polygonal curve associated with a track derived from the remote tracker to a corresponding polygonal curve on the planar map. A correspondence between the projected track and a track on the planar map is automatically established. A projection bias is removed based on the correspondence.

Abstract: Solutions for object tracking problems are presented by gathering images using one or more cameras, processing the gathered images to generate a directed acyclic graph, using the directed acyclic graph to determine a path cover that achieves maximum weight and satisfies one or more positive or negative constraints, and using the path cover to solve the object tracking problem. A first set of solutions utilizes trellis graphs, a second set of solutions employs a greedy approach, and a third set of solutions uses search algorithms.

Abstract: A system and method for projecting target tracks produced by a remote tracker onto a surface of the Earth to obtain projected tracks is provided. Projection bias in a projected track from the remote tracker projected onto a planar map is removed by computing a discrete Frechet distance from a polygonal curve associated with a track derived from the remote tracker to a corresponding polygonal curve on the planar map. A correspondence between the projected track and a track on the planar map is automatically established. A projection bias is removed based on the correspondence.