Abstract: Apparatus for generating a dynamic structured light pattern for optical tracking in three-dimensional space, comprises an array of lasers, such as a VCSEL laser array, to project light in a pattern into a three-dimensional space; and an optical element or elements arranged in cells. The cells are aligned with subsets of the laser array, and each cell individually applies a modulation, in particular an intensity modulation, to light from the laser or lasers of the subset, to provide a distinguishable and separately controllable part of the dynamic structured light pattern. A method of generating a structured light pattern is disclosed, in which light is provided from an array of lasers, and light is individually projected from subsets of the array of lasers to provide differentiated parts of the structured light pattern.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for neural network compression. In one aspect, a method comprises receiving a neural network and identifying a particular set of multiple weights of the neural network. Multiple anchor points are determined based on current values of the particular set of weights of the neural network. The neural network is trained by, at each of multiple training iterations, performing operations comprising adjusting the values of the particular set of weights by backpropagating gradients of a loss function. The loss function comprises a first loss function term based on a prediction accuracy of the neural network and a second loss function term based on a similarity of the current values of the particular set of weights to the anchor points. After training, the values of the particular set of weights are quantized based on the anchor points.

Abstract: Systems and methods for distilling deep neural networks are disclosed in which a teacher network is divided into blocks or “neighborhoods.” Candidate student models are then trained to reproduce the output of each teacher neighborhood, and the best student model corresponding to each neighborhood may be selected for inclusion in a final student network. In some examples, the final student network may be comprised of a collection of selected student models and copies of one or more teacher network neighborhoods.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining neural network architectures.

Abstract: A combination of two or more trained machine learning models can exhibit a combined accuracy greater than the accuracy of any one of the constituent models. However, this increase accuracy comes at additional computational cost. Cascades of machine learning models are provided herein that result in increased model accuracy and/or reduced model compute cost. These benefits are obtained by conditionally executing one or more of the models of the cascade based on the estimated correctness of already-executed models. The estimated correctness can be obtained as an additional output of the already-executed model(s) or could be determined as an entropy, maximum class probability, maximum class logit, or other function of the output(s) of the already-executed model(s). The expected computational cost of executing the model cascade is reduced by only executing the downstream model(s) when the upstream model(s) has resulted in an output whose accuracy is suspect.

Abstract: A technique for estimating the closest point of approach (CPA) of an object includes: tracking a range rate of the object based on Doppler measurements of the object; computing a first CPA estimate based on the tracked range rate; separately tracking a position of the object based on position measurements of the object; computing a second CPA estimate based on the tracked position; and computing a fused CPA estimate based on the first and second CPA estimates.

Abstract: A method for registering a radar system. The method includes obtaining first values for a location of a target relative to the radar system using radar system initiated signals, obtaining geo-referenced location data for and from the target, obtaining second values for the location of the target relative to the radar system using the geo-referenced location data, computing location registration bias errors for the radar system using the first and second values, and registering the radar system using the computed location registration bias errors.

Abstract: A receiver in a mode 5 air traffic control system provides amplitude and phase signal outputs a digital data stream containing preamble and flight information from data transmitted from an aircraft. A signal splitter divides the amplitude and phase signal outputs between an odd channel and an even channel that carry odd-numbered pulses and even-numbered pulses, respectively. An odd channel data decoder connected to the signal splitter extracts signals encoded in the odd channel and forms an odd data stream and an even channel data decoder connected to the signal splitter extracts signals encoded in the even channel and forming an even data stream. A preamble correlator correlates the odd and even data streams with a predefined preamble mask to detect potential valid preambles, and preamble decision logic processes signals output from the preamble correlator to identify which of the preambles actually are actually valid.

Abstract: A method for registering a radar system. The method includes obtaining first values for a location of a target relative to the radar system using radar system initiated signals, obtaining geo-referenced location data for and from the target, obtaining second values for the location of the target relative to the radar system using the geo-referenced location data, computing location registration bias errors for the radar system using the first and second values, and registering the radar system using the computed location registration bias errors.

Abstract: A receiver in a mode 5 air traffic control system provides amplitude and phase signal outputs a digital data stream containing preamble and flight information from data transmitted from an aircraft. A signal splitter divides the amplitude and phase signal outputs between an odd channel and an even channel that carry odd-numbered pulses and even-numbered pulses, respectively. An odd channel data decoder connected to the signal splitter extracts signals encoded in the odd channel and forms an odd data stream and an even channel data decoder connected to the signal splitter extracts signals encoded in the even channel and forming an even data stream. A preamble correlator correlates the odd and even data streams with a predefined preamble mask to detect potential valid preambles, and preamble decision logic processes signals output from the preamble correlator to identify which of the preambles actually are actually valid.

Abstract: A method for mitigating multipath impacts on azimuth accuracy in a monopulse interrogator is accomplished by calculating samples of monopulse ratio for samples of antenna boresight angles based on data received from an interrogation of a target. Samples of traditional target azimuth from the samples of monopulse ratio are calculated. A mean of the samples of traditional target azimuth is calculated. An alternative target azimuth from the samples of monopulse ratio is calculated. Whether a multipath signal exists is determined from observing a standard deviation of the samples of traditional target azimuth, and using the mean of the traditional target azimuth if a multipath signal does not exist and using the alternative target azimuth if a multipath signal does exist.

Abstract: In a radar system, a monopulse calibration table is constructed from live targets of opportunity. A center of gravity or weighted average of normalized signals ?V received at SUM and DIFF channels from a live target are used to determine the target's actual azimuth. Off bore sight angles (OBA) of the target are then determined from the target's actual azimuth. Normalized received signal values of ?V are converted to nearest-valued integers. The OBA s that correspond to each integer-valued normalized received signal are averaged and can then be plotted as a function of normalized received signal value ?V. Different tables or plots can be constructed for elevation angles. An equation of a best-fit line the matches or at least closely approximates the plotted data is determined to smooth the actual data.

Abstract: A method for mitigating multipath impacts on azimuth accuracy in a monopulse interrogator is accomplished by calculating samples of monopulse ratio for samples of antenna boresight angles based on data received from an interrogation of a target. Samples of traditional target azimuth from the samples of monopulse ratio are calculated. A mean of the samples of traditional target azimuth is calculated. An alternative target azimuth from the samples of monopulse ratio is calculated. Whether a multipath signal exists is determined from observing a standard deviation of the samples of traditional target azimuth, and using the mean of the traditional target azimuth if a multipath signal does not exist and using the alternative target azimuth if a multipath signal does exist.

Abstract: In a radar system, a monopulse calibration table is constructed from live targets of opportunity. A center of gravity or weighted average of normalized signals ?V received at SUM and DIFF channels from a live target are used to determine the target's actual azimuth. Off bore sight angles (OBA) of the target are then determined from the target's actual azimuth. Normalized received signal values of ?V are converted to nearest-valued integers. The OBA s that correspond to each integer-valued normalized received signal are averaged and can then be plotted as a function of normalized received signal value ?V. Different tables or plots can be constructed for elevation angles. An equation of a best-fit line the matches or at least closely approximates the plotted data is determined to smooth the actual data.

Abstract: A heat sink (100) includes a cooling block (105) having a plurality of fins (110) extending through the block to provide a series of coolant channels (112) extending between the fins, the channels having in the interior of the block a venturi narrowing. Beneficially, the channels have coolant entry regions in at least two different regions (115, 120) of the block to direct a plurality of coolant flows into the interior of the block from different directions. Also described is a method of cooling a device such as a processor of a mobile computer used in a vehicle by placing the heat sink or more than one of the heat sinks adjacent to the device to be cooled, preferably with a region in which the venturi narrowing of the coolant channels is provided close to the hottest part of the device to be cooled.

Abstract: An exemplary radar system includes a waveform generator that generates a control waveform. An in-phase and quadrature modulator receives the control waveform from the waveform generator and in turn generates a waveform output that is amplified by a power amplifier before being transmitted from an antenna.

Abstract: A heat sink (100) includes a cooling block (105) having a plurality of fins (110) extending through the block to provide a series of coolant channels (112) extending between the fins, the channels having in the interior of the block a venturi narrowing. Beneficially, the channels have coolant entry regions in at least two different regions (115, 120) of the block to direct a plurality of coolant flows into the interior of the block from different directions. Also described is a method of cooling a device such as a processor of a mobile computer used in a vehicle by placing the heat sink or more than one of the heat sinks adjacent to the device to be cooled, preferably with a region in which the venturi narrowing of the coolant channels is provided close to the hottest part of the device to be cooled.

Abstract: A radar based application programmable waveform generator component of an apparatus generates a waveform and controls a phase and an amplitude of one or more parts of the waveform.

Abstract: A method for eliminating signals from false targets in a secondary surveillance radar system includes the steps of obtaining plots of radar data for the same target. The range from the radar system to the targets is determined, and the target plot having smallest range of all target plots that indicate reflection of radar signals from the same target is accepted as being a real target. Other target data is rejected as being from a radar reflector. The coordinates of radar reflectors may be stored in a database to form a map of reflectors in the vicinity of the radar.

Abstract: A method for eliminating signals from false targets in a secondary surveillance radar system includes the steps of obtaining plots of radar data for the same target. The range from the radar system to the targets is determined, and the target plot having smallest range of all target plots that indicate reflection of radar signals from the same target is accepted as being a real target. Other target data is rejected as being from a radar reflector. The coordinates of radar reflectors may be stored in a database to form a map of reflectors in the vicinity of the radar.