Research Projects

The SAM Lab develops signal processing methods to extract information from multidimensional data collected by acoustic or electromagnetic sensors. The methods are used in fields of sonar, radar, biomedical imaging, wireless communications, and many more. By using knowledge of wave propagation to in complex media, we design information extraction techniques that work in complex media.

MIMO radar for vehicular situational awareness

Radar systems are most commonly deployed on fixed platforms, aircraft, or marine vessels. Radars are often precluded from use on ground vehicles due to high amounts of multipath reflections off of clutter in the environment. However, using adaptive MIMO radar, the signal processor can perform "virtual non-causal transmit beamforming," which effectively steers the transmit beam away from strong scatterers in the environment and filters out multipath clutter returns.

MIMO Array Calibration from Clutter

Calibration of radar systems is often a long, tedious, and inexact task, which often involves mounting multiple calibration sources and making a copious amount of measurements. In many circumstances, it is not practical to mount calibration sources, which makes it very difficult, if not impossible, to properly calibrate the radar.

However, MIMO radar signal processing can be used to calibrate a radar without any calibration sources. Since MIMO radar allows for after-the-fact beamforming on both the transmit and receive sides, environmental clutter can be used as a pseudo calibration source by steering the transmit beam to illuminate a specific clutter patch.

Similarly, the transmitter can be calibrated by steering the receive beam. By iteratively alternative between the transmit and receive beamformers, a calibration solution can be converged on. A significant advantage of this method is that all of the processing is done after-the-fact, and therefore very few measurements need to be taken..

Horizontal Towed Array Shape Estimation

When hydrophone arrays are pulled behind ships in the ocean, the array will bend as the ship turns or maneuvers. When the array shape is unknown, the performance of sonar systems degrades. Typically, towed arrays have heading sensors in order to estimate the array shape. However, during sharp turns or due to the rough operating conditions of the open ocean, heading sensors can temporarily or permanently fail. The shape of the array can be estimated by exploiting the noise field directionality. Distant ships and ambient ocean noise are exploited to maintain sonar detection and localization performance during tow platform maneuvers. Additionally, the hydrophone data and heading sensor data can be fused to increase robustness.

Geo-location in Dense Multipath

We are also interested in the problem of target geo-location in dense multipath environment such as indoor and urban area. Our solution is based on jointly estimation of target position and multipath channel parameters, referring as Simultaneous Target and Multipath Positioning (STAMP). We consider both monostatic and multistatic localization case with hybrid angle of arrival (AOA)/ angle of departure (AOD)/time of arrival (TOA) observations, and the Bayesian Cramer-Rao Lower Bound (CRLB) is derived and analyzed for various multipath propagation models in terms of information gain and loss. Practical implementation issue such as data association uncertainty is resolved by a multi-hypothesis data association scheme incorporating with Extended Kalman Filter (EKF).

Continuous Active Sonar Waveform Design

Conventional Pulsed Active Sonars (PASs) use short, high energy pulses or Linear Frequency Modulated (LFM) chirps to highlight targets with long delays (e.g. 10 sec) between consecutive transmissions. The long delays are used to mitigate “ghost targets” known as range ambiguities at the cost of making the sonar operator wait for each range and velocity update. For the CAS waveform project, we have designed a waveform that can be transmitted continuously (no delays) that uses LFM chirps with frequency staggering based on circular Costas sequences. Circular Costas sequences have a special property that allows us to mitigate range ambiguities while enabling target range and velocity estimates every 0.2 seconds. The figure shows a cartoon version how a waveform with frequency staggering based on the Costas sequence (4, 7, 1, 6, 5, 2, 3) leads to a nearly ideal thumbtack ambiguity function.

Human Motion Monitoring with In-room Radar

Multi-Input-Multi-Output (MIMO) systems have been successfully used for mitigating multipath propagation having different Direction-of-Arrival (DoA) than Direction-of-Departure (DoD). In this project, a MIMO RF probe is used for monitoring multiple moving targets simultaneously in an indoor environment such as a hospital ward. The Range-Angle map indicating the locations of moving targets is much cleaner using MIMO processing, which demonstrates the ability of MIMO to mitigate the effects of multipath.