Automotive Sensors: Steering The Development Of Autonomous Vehicles
By Emily Newton
Autonomous vehicles are some of the most exciting technological developments today. While there’s still a long way to go before fully self-driving cars are a reality, researchers are getting closer every day. Advances in automotive proximity sensors are the key driver of this progress.
Getting a car to drive itself requires advanced machine learning models, but these algorithms are only as useful as the data they interpret. Consequently, sensors must provide as much precise, reliable information as possible for the self-driving AI to understand its surroundings.
Types Of Automotive Proximity Sensors
The sensors within autonomous cars vary widely, and many automakers combine many into a single system. While manufacturers may disagree on which proximity technology holds the most potential, all categories have undergone crucial improvements in recent years.
LiDAR
Light detection and ranging (LiDAR) is one of the most prominent automotive proximity sensor technologies. Using lasers to map an area is not new in and of itself, but the technology has made several strides within the realm of autonomous vehicles.
Frequency-modulated continuous-wave (FMCW) lidar measures the frequency of the outgoing signal and incoming reflections. Calculating the difference allows it to measure velocity and trajectory, not just position. It’s also more resistant to interference and provides greater detail.
Advances in micromaterials take FMCW lidar’s benefits further. Micron-scale silicon systems have enabled some to achieve a range of 120 meters with 0.3-meter precision. Relying on microscopic electronics also leads to smaller, lighter lidar solutions to free space within the vehicle.
Radar
Radar is a similar solution. Radar is an older technology already common in cars, but it’s typically less reliable than lidar. Advances in photonics may change that.
Photonic radar still relies on radio waves to map an area, but it analyzes the signal with lasers to boost their range and precision. Alternatively, systems could use fiber optics to send and process information instead of conventional electric wires. In both cases, the introduction of photonics makes radar a far better solution than it used to be.
Technologies like this have achieved resolutions as precise as 1.8 centimeters while transmitting data at 100 megabits per second. In a self-driving car, such an improvement could dramatically improve the vehicle’s ability to navigate safely without the extra expense of lidar.
Cameras
Some automotive proximity sensors mimic the human eye instead of measuring distances with radio or light waves. Machine vision allows self-driving algorithms to interpret visual data from onboard cameras. While it may not offer precise measurements, it can recognize obstacles with impressive clarity.
Tesla famously solely relies on optical sensors, forgoing radar or lidar entirely. Doing so allows the company to keep costs down, making cars with autonomous features accessible to a wider range of drivers. As machine vision progresses, its potential could go even further.
Color-depth (RGBD) cameras — which can measure depth, position, and speed through a stereo camera setup — are leading the charge. RGBD offers higher image density than lidar or radar at close distances, helping it detect nearby obstacles and differentiate between different types with higher accuracy.
Photoelectric Sensors
Smaller-scale autonomous driving features may rely on photoelectric sensors. There are two main types of photoelectric sensors — those using a single device to send and receive signals and systems with separate emitters and detectors. In either case, these simpler solutions handle close-range proximity detection for precise movement.
Parking sensors and blindside monitors are among the most prevalent applications. Ranges and resolutions tend to be lower with photoelectric options, but they offer one crucial advantage — they’re cheap.
Advances in electromagnetic and laser technology have led to remarkably compact, low-cost photoelectric proximity sensors. As a result, cars can reduce the costs and complexity of parking or lane detection systems to leave room in the budget for more sophisticated driving tasks.
What’s Next For Automotive Proximity Sensors
Across all proximity detection technologies, a few obstacles remain. Reliability is a chief concern in every solution. Cameras require proper lighting to be accurate, radar usually has low resolution and photoelectric sensors cannot distinguish between surfaces. Laser-spoofing techniques can trick lidar systems into seeing false obstacles with a worrying 92.7% success rate, according to recent research.
Sensor fusion may be the optimal solution. While each technology has its own security vulnerabilities and accuracy shortcomings, these weaknesses don’t always cross over. Self-driving AI could get a better picture of its surroundings by analyzing data from multiple systems, ensuring safe navigation even when one sensor fails.
Costs are another common concern, especially with high-end lidar and photonic radar solutions. Thankfully, development has been trending in a positive direction in this area. Average lidar prices have fallen from $10,000 to around $500 within the past decade. Expenses will keep falling as production scales up and research finds new ways to manufacture automotive proximity sensors.
Autonomous Cars Need Next-Generation Sensors
Humans can’t drive safely without clear vision. In the same way, autonomous cars can’t navigate without reliable and versatile sensing technologies. Consequently, the future of self-driving vehicles lies with the organizations developing the next generation of lidar, radar, machine vision, and photoelectric sensors.
Significant obstacles remain, but these technologies are moving in the right direction. Before long, they could reach the point where level 5 autonomy becomes a reality.