SuperRadar - A New Dimension in Automotive Radar Perfomance
Today’s vehicles must be able to sense and interpret their surroundings to then make correct, split-second decisions based on the environmental model.
Symeo has invented a revolutionary Radar signal processing technique called SuperRadar that enables coherent processing of multiple automotive Radar sensors to increase the performance of a given Radar system along multiple dimensions without the need for additional HW, but by adding SW only.
In a conventional automotive Radar system, each Radar operates independently and receives and processes only its own transmit signals or direct paths. SuperRadar enables Radar sensors to additionally receive and process signals that were transmitted by other radar sensors to yield the so-called cross paths. To align in time, the Radar sensors are pre-synchronized via a common clock, that is distributed over a conventional data bus and triggers the transceivers of all radar sensors simultaneously. The receive signals of the cooperating Radar sensors are transmitted to a central compute platform where frequency and phase errors are corrected by the SuperRadar algorithm and the SuperRadar point-cloud is generated.
SuperRadar Working Principle
For verification purposes, Analog Devices has implemented the technology on a test vehicle. The vehicle is equipped with 2 cooperating 77 GHz Radar sensors that use FMCW modulation and multiple transmit and receive antennas to interrogate their environment in 3D.
Applying SuperRadar to the overlapping field of view of multiple Radar sensors offers several key features.
As the SuperRadar algorithm allows the capture of the cross-path reflections in addition to the direct path reflections, more reflection points are generated in the environment. Those additional reflection points increase the detection probability, provide more information about the interrogated targets and hence increase the reliability and availability of the Radar systems output data. The additional reflection points can be used for improving object classification and grid map generation and help with detection of targets with low or angle dependent Radar Cross Section.
Additional Reflection Points
The SuperRadar image is generated by processing twice the amount of MIMO channels compared to a single sensor, which improves the angular resolution of the SuperRadar image by a factor of up to 2. A better angular resolution helps with separation of multiple targets that are side-by-side especially at higher distances and when small radar cross section targets such as vulnerable road users are close to targets with high radar cross sections. If more than 2 Radar sensors are used cooperatively, the resolution can even be increased to a level that resembles the performance of Lidar sensors.
Enhanced Angular Resolution
The horizontal displacement of transmit and receive channels enables estimation of the full velocity vector of all targets within a single Radar burst which can be used to estimate target velocities with significantly lower latency compared to a conventional Radar system. This enables much faster reaction times and reduces the risk of overreaction of advanced driver assistance and autonomous driving systems when encountering dynamically moving road users. The tangential doppler component also enables detection and classification of tangentially moving objects with much higher reliability compared to conventional radar.
The following Videos show selected scenes from Test drives with our vehicle: