Autonomous Vehicle Radar: Improving Radar Performance with Simulation

By Shawn Carpenter, Product Manager, High Frequency Electronics, ANSYS

Radar systems provide important sensor input for safe and reliable autonomous vehicle operations. Ensuring that these radar systems operate without interference, cover the intended areas, do not fail from installation effects and provide accurate input to the control system requires use of advanced engineering simulation.

Save PDF
Autonomous Vehicle Radar

automotive sensors

car sensors 

Autonomous vehicles require the continued evolution of vehicle sensors — the eyes and ears of the control system that perceive the operational characteristics of the vehicle and the environment around it. The sensors feed the vehicle control systems with data on the current and developing state of the vehicle’s surroundings. Both operation and safety depend on the accuracy of the sensor system.

Four major classes of vehicular sensors provide the lion’s share of environment sensory data for an autonomous vehicle — visual spectrum cameras, laser-ranging devices (lidars), ultrasound sensors and radio frequency ranging sensors (radar). Automotive radar employs millimeter-wave frequencies for long-range object and obstacle detection, as well as for tracking the velocity and direction of the various actors such as pedestrians, other vehicles, guardrails, etc., in the environment around the vehicle.

SIMULATION OF A RADAR ANTENNA

This example illustrates the development of a 77 GHz automotive radar sensor based upon a two-sided printed circuit board (PCB) fabrication technique using a slotted waveguide.

slotted waveguide unit cell

ANSYS HFSS is used to quickly optimize proper dimensions for each element of a slotted waveguide unit cell in an extended antenna array.

waveguide array

Once the design of the unit radiation cell is optimized for the desired frequency, a full-size array can be laid out quickly and automatically. Simulations are run with an automated and highly scalable technique to determine the minimum number of cells required to achieve spatial radiation coverage and the efficiency with which the array radiates power.

array simulation

When an array design satisfies performance requirements, the fabrication details (vias, metal thicknesses, structures to couple power into the waveguides, etc.) can be added to simulate realistic materials and manufacturing processes. A digital exploration design of experiments (DoE) can be run against the expected fabrication process tolerances to assess the manufacturing yield of this array. This initial design of the array with vias, PCB filler and transitions shows a simulated far-field radiation pattern when all the array elements are fed with power.

packaging simulation

The effects of the packaging and housing can be investigated to understand their influence on the sensor’s performance. Metal in and near the packaging can create electromagnetic coupling to the array that might degrade its ability to radiate to specification. Proximity effects of the radome and other nonmetallic packaging can also have an impact. These effects can be determined and even remediated in a simulation model prior to building a physical prototype.

packaging environmental effects

Environmental effects on the packaged sensor’s performance can also be considered, such as rain, ice, dust or other materials. In this simulation, a thin layer (0.1 mm) of water or ice is studied over the radar package, showing that the water has minimal effect on the main beam gain, but increases the sidelobe level by another 4 dB. By understanding performance under different environmental conditions, engineers can optimize the array’s design and build appropriate margins into the original design.

"For automakers to gain the full benefits from automotive radar technology they must judiciously use simulation to meet development schedules and to achieve performance requirements."

Three major classes of radar systems are typically employed in automotive active safety systems:

  • Short-range radar (SRR) for collision proximity warning and safety, and to support limited parking assist features
  • Medium-range radar (MRR) to watch the corners of the vehicle, perform blind spot detection, observe other-vehicle lane crossover and avoid side/corner collisions.
  • Long-range radar (LRR) for forward-looking sensors, adaptive cruise control (ACC) and early collision detection functions.

Today’s automotive radars incorporate technology that 20 years ago could only be found in advanced research in aerospace and defense laboratories. For automakers to gain the full benefits of this technology — including chip-level integration, package and sensor miniaturization, fewer parts, lower power consumption, and higher performance, all at dramatically lower costs — they must judiciously use modeling and simulation to meet aggressive development schedules and achieve challenging performance requirements.

Radar simulation can be employed to design single radar components (antenna and array), develop a system including all radar installations and the vehicle, or even extend to a virtual system of multiple radar systems, the vehicle itself and its environment — a digital prototype.

Rapid Development of Radar Sensors

High-performance radar design starts with the antenna — the interface between the sensor and the world that it is sensing. Ideally, these antenna systems must concentrate energy in one direction over a defined coverage angle. Antennas must radiate efficiently so that energy is not dissipated in the antennas themselves or in the sensor package materials. Energy should not be lost due to poor match with the transmit power amplifiers.

High-frequency modeling and simulation present tremendous opportunities for time and cost savings in the design and development of radar sensors. With simulation, engineers can:

  • Virtually prototype and “tune” antenna topologies quickly, without requiring fabrication.
  • Test antenna variants effectively and efficiently to understand their behavior under a variety of structural and environmental conditions
  • Optimize element and multichannel antenna arrays with the least effort and cost.
  • Build only a single prototype to test at the end.

Integrating the Radar with the Vehicle

Once a sensor design or prototype is developed, it must be evaluated as installed on a vehicle. Many radar sensors are mounted either behind a bumper or in the vehicle fascia. The proximity effects of the vehicle design can affect the performance of the radar — particularly the antenna’s ability to focus radar energy. The vehicle manufacturer develops bumper and fascia designs to be both aerodynamic and aesthetically pleasing to their buyers. The unique features of a body shape that meet aesthetic goals could negatively impact the performance of a radar sensor integrated into it or hidden behind it.

In the past, the effects of radar-to-fascia and radar-to-bumper interaction were evaluated through cooperation between the sensor manufacturer and the vehicle manufacturer. This was an iterative process based on trial-and-error prototyping. Valuable development time and cost were invested in prototypes that required retooling as the car was redesigned.

Modeling and simulation reduces this process from as long as nine months to a matter of days. ANSYS HFSS SBR+ can integrate models, including highly accurate results from finite element ANSYS HFSS models, for the isolated sensor system, and simulate its interaction with the much larger fascia and bumper using its high-frequency ray tracing methods. The simulated installed radar antenna response shows the radar engineer how each radar subarray will illuminate the road or environment when it is installed into the proposed fascia–bumper design.

"Radar systems play a central role in safety systems and must be tested with vehicle control systems and algorithms to validate safe operation"

radar sensor array
A radar sensor array model installed in a proposed automobile fascia
radar sensor array simulation
The ANSYS HFSS SBR+ shooting and bouncing rays EM field solver is applied to model the installed radar sensor array interactions. The HFSS finite-element simulation for the radar sensor antenna system is shown in the proper installation location, and a subset of rays employed by the HFSS SBR+ simulation is shown at an exit angle of 80 degrees.

Virtual Road Testing for Radar

Autonomous vehicle developers are devoted to the safety of passengers. Radar systems play a central role in safety systems and must be tested with vehicle control systems and algorithms to validate safe operation. Without the benefit of modeling and simulation, this would require driving millions of test miles. Today, most AV developers are moving this process to the domain of the digital prototype. In modeling and simulation, testing can be performed for any conceivable scenario.

However, high-fidelity modeling of the electromagnetic performance of an automotive radar system has to this point proven to be a major challenge.

Full-physics modeling and simulation of radar sensors creates an enormous EM analysis problem as the radar needs to cover an area that could occupy over 1.4 million electrical wavelengths. This is compounded by system-level requirements that include the number of times the central control system is updated by radar, the number of antennas involved, the range and velocity resolution of the MRR system, and the comparative velocity of the environmental actors.

While these considerations pose challenges to high-fidelity EM modeling of radar–environment interaction, they are not insurmountable. An appropriate application of the shooting andbouncing rays (SBR) technique using ANSYS HFSS SBR+ can provide full-physics simulation of such problems with good accuracy and reasonable efficiency — in terms of both computer resources and modeling time.

Tx channel sensor simulation
Rx channel sensor simulation
ANSYS HFSS–simulated near-fields surrounding sensor for Tx channel and for Rx channel form the basis for excitation in the HFSS SBR+ fascia interaction solution.
Receive channel subarray radiation pattern
Receive channel subarray radiation pattern shows radiation patterns for the module in isolation, and as installed to include fascia and bumper interaction.
transmit channel radiation pattern
Transmit channel radiation pattern shows radiation patterns for the module in isolation, and as installed to include fascia and bumper interaction.

ANSYS HFSS SBR+ can be used to synthetically reproduce the signals obtained by a high-fidelity radar model. Any specified bandwidth may be applied to the simulation to foster virtual innovation by enabling the engineer to test new waveforms that may not be currently available from sensor suppliers.

Radar signal processing systems need to intelligently group distributed target returns that belong to the same actor in the environment, or the vehicle control system will be overwhelmed with too many targets to track. This grouping is made possible by processing the possible Doppler-shift of the signals that bounce off surfaces with a velocity that is different from the observation domain. Radar signals from targets that have the same velocity in consecutive range bins can be considered to be from the same target. Accurate determination of targets in terms of both range and velocity requires a large number of pulses to be analyzed over time. ANSYS HFSS SBR+ results make it possible to develop Range-Doppler maps that show the range to the target returns on one axis and the extracted velocity of the targets on the other.

A typical automotive radar sensor provides updates to the vehicle control and safety systems at a rate of 5 to 30 frames per second. The speed and accuracy of ANSYS HFSS SBR+ allows a complete simulation of the movement of vehicles through this environment to develop a Range-Doppler map over time for the scene. Placing the HFSS SBR+ simulation within the simulation loop of a complete autonomous vehicle creates a digital prototype to test the vehicle control system or active safety system.

intersection simulation
Busy intersection environment geometry. Velocity of each moving actor in the scene is shown.
SBR
Shooting and bouncing rays traced from radar transmit channel throughout the environment. Multiple colors correspond to ordinal reflection for each ray track pictured.
Range-Doppler map
Range-Doppler map for radar system over a radar frame of 200 consecutive 300 MHz pulses

Complete Modeling and Simulation Work Flows

Radar sensor developers, automotive OEMs, active safety systems developers and autonomous vehicle control systems developers use ANSYS solutions to design radar sensor modules, study their installed performance on the vehicle, and gain insight into radar reports for moving and stationary targets on a full, dynamic road scene. From a single component to a digital system prototype, ANSYS provides unique solutions for this very challenging high-frequency problem.

radar range profile

Range profile for a single radar 300 MHz bandwidth pulse with 0.5 m resolution. Radar is visible in overlay at lower left. The range profile shows distance of flight for all radar echoes received by the radar system in response to the modeled environment. Very strong radar returns for some light posts, the surfaces of several of the vehicles, and between vehicle reflections are shown. The signals are stronger from closer targets than from more distant targets, but even targets well down the road are detectable. Due to the waveform’s resolution, the radar may detect multiple target echoes from the same vehicle.

"Radar systems play a central role in safety systems and must be tested with vehicle control systems and algorithms to validate safe operation."

radar system

click below to start a conversation with ANSYS

Contact Us
Contact Us
Contact