According to a 2015 dossier published by the U.S. National Highway Traffic Safety Administration (NHTSA) on traffic safety, 94% of accidents were caused by human (read driver) error. Advanced Driver Assistance Systems (ADAS) employ several radars around the vehicle that help detect and localize the environment around the vehicle.
Currently, radars are based on the 24GHz or the 77GHz bandwidth, with suppliers focusing on migrating to 77GHz for all applications since they support improved accuracy in distance, superior speed measurements, and more precise angular resolution. Furthermore, the antenna size of a 77GHz is smaller with lower interference issues.
As we steer into an ecosystem with autonomous vehicles at the helm, it is important that the industry recognizes some of the red flags associated with using radar systems. Object classification and recognition, along with the difficulty in detecting small objects such as a bicyclist or a lamp post, are probably the biggest drawbacks of using radars in automotive applications. Despite their shortcomings, radars are fast becoming the base sensor for OEMs and suppliers to push various ADAS functions at a relatively economical pricing.
Next Generation Radar Sensors
As stated earlier, due to the inherent drawbacks of radars, it is difficult for OEMs to have just radar sensors for ADAS. For this reason OEMs have been employing a sensor suite comprising cameras, radars, and LiDAR. Only the camera and LiDAR have the capability to identify objects around the vehicle. However, both are incapable of operating in excess of 300m at high speeds. This poses a challenge for vehicles driving at higher speeds. On the other hand, radars can operate in excess of 300m even at higher speeds and can easily function in the worst of weather conditions. However, object classification in the case of radars is not possible.
These intrinsic weaknesses have forced OEMs, suppliers and other market incumbents in the radar space to focus on developing radars that are capable of identifying objects. Next generation ultra-high resolution imaging radars are quickly becoming the most sought after technology for autonomous driving. 4D radars offer real-time object detection which works in all and any weather and lighting conditions.
Unlike current generation radars, ultra-high resolution imaging radars provide detailed imagery of the surroundings with a wide field of view (FoV). This means that not only are objects around the vehicle detected but so are objects like pedestrians or motorcycles that might be masked by larger objects such as a tree or truck. Moreover, these radars are capable of determining whether the object is stationary or in motion, and in which direction they are moving, while providing the vehicle with real-time data with ranges exceeding 300m.
How it differs from current generation radars
Today’s radar systems play a pivotal role in ADAS features such as adaptive cruise control, blind spot detection, and automated emergency braking. However, radar suppliers have to make a functional choice between medium resolution with a limited FoV or low resolution with a wider FoV. To achieve Level 4 and 5 autonomy in the future, it is imperative that incumbents in the radar domain make use of high resolution radars capable of sensing the environment at a wide angle of 100⁰ FoV.
Ultra-high definition imaging radars are also able to provide path planning features as they can create a vivid image of the road ahead at a distance far beyond 300m, and capture the size and velocity of an object in front. A special focus on object separation by elevation enables the radar to recognize whether the vehicle will encounter a stationary object right in front of it and must stop or a bridge under which it can safely drive.
Additionally, an important parameter to achieve Level 4 and 5 autonomy is to identify and filter out false positives. A 4D imaging radar is capable of this ability; it can filter out false alarms, while providing optimal sensitivity. The radar uses the lowest detection threshold, so as to report even the faintest noise. Post processing and tracking are used to filter out random noise, while calibration schemes allow extremely low side lobe levels to be reached.
Will next generation 4D radars eventually diminish the value of LiDAR
Radar systems over the years have gone through a series of incremental additions in order to improve their performance and capabilities. For instance spatial resolution of radars have improved to 0.1 degrees which is comparable to some of the high resolution LiDAR’s available in the market; albeit at a fraction of a cost. This was achieved by increasing the number of transmitter and receiver antennas to form a virtual array using multiple-input and multiple-output (MIMO) signal processing. Additionally, enhanced Frequency Modulated Continuous Wave (FMCW) modulation techniques have helped radar systems to discriminate objects in highly dense and ambiguous scenarios, a function which they were earlier incapable of doing. Furthermore, down conversion of radio-frequency modulation sample rates to beyond 40 MegaSamples per second to aid long Fast Fourier Transforms (FFT) up to 4,000 points have allowed for the range span to be divided into smaller cells to reveal hidden details that previously tended to be missed.
Such enhancements to the eventual output of radar systems, coupled with 4D radars that allow for the detection of range, speed, azimuth and elevation, have eroded some of the key advantages of LiDAR. Moreover, when 4D radars are combined with multi-ocular cameras, it is technically possible to completely eliminate the need for a LiDAR. This can be seen from Tesla’s approach to L4 and L5 autonomy.
Radar is a proven technology that is becoming increasingly more efficient at sensing a vehicle’s surroundings. The recent slew of radar launches using Radio Frequency Complementary Metal-Oxide Semiconductor (RF CMOS) technology by the automotive industry will allow for smaller, lower power and more efficient radar sensors that fit into OEMs’ cost reduction strategies. Moreover, the next generation of radars will make autonomous vehicles exponentially safer by being able to detect objects in excess of 300m. With their 4D imaging capabilities, radars could eventually replace cameras; however, for that to happen, radars must be able to “see” color, which they currently cannot.
The automotive radar market has never been so dynamic. As we enter an exciting era of technological innovation, new opportunities for radar are still emerging. These include vital-sign driver monitoring systems, chassis-to-ground monitoring, and hands-free trunk opening, among others. The development of autonomous vehicles has reached a crossroad. In response to recent concerns, the advanced mobility industry needs to revisit the role of the high-resolution imaging radar as an indispensable element in the autonomous sensor suite. There’s no doubt this technology will be key in autonomous and robotic cars.