Safety First: 4 Trends Inspiring Innovation in Autonomous Vehicles

Safety First: 4 Trends Inspiring Innovation in Autonomous Vehicles When it comes to driving, safety is top priority. However, no matter how safe a vehicle is, human error can complicate things. The mistakes drivers make behind the wheel account for 94% of all vehicle accidents. In 2019 alone, 6.8 million vehicle accidents¹ were reported in the U.S. with 1.9 million accidents resulting in injuries and fatalities. The National Highway Traffic Safety Administration (NHTSA) estimates that vehicle crashes in the U.S. result in more than $836 billion in economic cost and quality of life each and every year. To mitigate risk, automotive manufacturers are introducing increasingly sophisticated automated driving systems in vehicles, with an emphasis on improving safety and delivering economic benefits. These systems aim to eliminate human error, thereby making traveling in vehicles safer – but there’s still a way to go. Autonomous Driving Can Be Safer, but Is It Viable? In 2014, The Society of Automotive Engineers defined six levels of driving automation in its SAE J3016™ Recommended Practice: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles↗, commonly referred to as the SAE Levels of Driving Automation™. The levels range from 0 (no automation) to 5 (full automation). The SAE defines the↗ five levels of automation as follows:

Currently, none of the cars² on U.S. roads are fully automated. The highest level autonomous systems commercially available today are L2 where the driver must pay attention to the road and have at least one hand touching the steering wheel. Still, continued innovations from vehicle OEMs are inspiring, and as more automation is introduced, driving will become increasingly safer. Case in point: The Insurance Institute for Highway Safety (IIHS) reported a 7% reduction in crashes³ for vehicles with a basic forward-collision warning system, and vehicles with automatic braking saw a 14-15% reduction in crashes. While fully self-driving vehicles that operate without the intervention of a human driver dominate the discussion about autonomous vehicles (AVs), it will likely take another 10 years before these vehicles become mainstream and capable of delivering people in private passenger vehicles from point A to point B. The delay in adoption is due in large part to their high cost. Today, AVs deployed in test fleets cost upwards of $180,000 – a price that’s prohibitive to widespread adoption. Costs will inevitably drop, but it will take a while. The initial roll-out for these systems will be for commercial applications, where revenue from services delivered can be used to offset the cost. For example, much work has been invested in developing robo-taxis↗ for mobility as a service (MaaS), as well as AVs for long-haul trucking↗ and last-mile delivery↗ applications. But momentum is indeed ramping up to prepare for a future that includes AVs – in June last year, Waymo, creator of Google’s self-driving car↗, announced $2.5 billion in additional funding⁴ to be used toward advancing its autonomous driving technology. Safety-First Approach will Drive Innovation As the industry works toward bringing AVs into the mainstream, more safety features for private passenger vehicles will be developed and introduced. Here are four development trends we expect will gain momentum:

1) Growth of ADAS Systems: Typically referred to as advanced driver-assistance systems (ADAS), L2 and lower levels of autonomous driving feature electronic systems designed to increase the safety of a vehicle during operation. It’s also common today for vehicles to be equipped with driver assistance or partial automation features, such as lane-departure warning, forward-collision warning, adaptive cruise control, and lane change assistance. L2 entry and advanced systems are expected to be present in 59% of all new vehicles by 2025⁵, and the market for ADAS systems will reach $74.9 billion by 2030⁶. 2) Rise of partial hands-free operation: L3 is the first-level of autonomous driving which operates under specific conditions. Over the next few years, OEMs are planning to release vehicles with L3 systems, operating over predefined highways or locations. Known as “geofencing⁷,” the vehicle defines spatial boundaries and references detailed maps of the surrounding terrain. The car projects sensor data onto the maps to determine the safest route. As with L2 systems, drivers must be ready to take control of the vehicle at all times. Examples of partial hands-free operation are GM’s Super Cruise↗ or Ford’s BlueCruise↗. L3 systems that feature this level of automation will begin to outpace L2 systems in the latter half of the decade. 3) More sensors and increased resolution: When sensors first made their appearance on the automotive scene, they were expensive – but new technology becomes more affordable over time. While today's vehicles already use intelligent sensors for a variety of functions – controlling and processing oil pressure, temperature, emission and coolant levels, to name a few – as the cost of sensors decreases, manufacturers will continue to incorporate more of them to enable new features such as full 360-degree visibility. Additionally, subsequent generations of sensors will provide increased resolution. Whereas most camera sensors today offer 1-2 megapixels, future camera sensors will provide resolutions of close to 8 megapixels, for higher-quality images and video that improve the ability to identify objects in the environment. The combination of camera, RADAR and LiDAR sensors operating at the same time ensures reliability through redundancy. For example, if a bright light renders a camera sensor inoperative, data from another sensor such as RADAR will ensure continuous operation. A data aggregation system collects data from all the sensors and combines it to generate a comprehensive “picture” of the vehicle's surroundings. This operation, referred to as sensor fusion, is performed by a high-performance SOC or FPGA that processes the data and hands it off to the ADAS/AD system, for interpretation and to aid in real-time decision making. ● More sophisticated AI-driven systems: As the number of autonomous vehicles operating on roads increases, the more data we’ll have at our disposal. Data from the camera’s sensors can be uploaded to the cloud, where it can be analyzed and used through machine learning to train new, more sophisticated AI inferences. These inferences are then sent back to the vehicle via over-the-air (OTA) updates, initiating a continuous loop of refinement. Over time, a vehicle’s responses to real-world driving situations and “corner-cases” – problems that occur outside normal operating parameters – will continue to improve.

Storage, Memory and Power Management Are Essential for Success

All of these trends hinge on the availability of sufficient processing power, storage and memory being built into the ADAS and AD systems. “SoCs powering these systems need ever-faster I/O performance for continuously-evolving AI inferences to respond to traffic situations in real time,” said Ryan Suzuki, Sr. Product Marketing Manager at Samsung. “In addition, multiple and higher-resolution sensors will require SSDs to support higher capacities as well as faster sequential write operations to store images and video.” Samsung is working to manufacture the memory required to support these systems today and higher-performance memory for future next-generation systems. We recently introduced new automotive solutions aimed at advanced infotainment and autonomous driving systems – high-performance SSDs, graphics DRAM, LPDDR4/5 and UFS products. Read up on our solutions for automotive to learn more about how Samsung is fueling the future of safer driving.

---

_{¹ In 2019, nearly 6.8 million vehicle crashes occurred in the United States, around 1.9 million of which were injury crashes. Roughly five million vehicle crashes, a little more than 70 percent, only involved property damage. ² At the moment, no self-driving car operating on US roads is completely autonomous. ³ The Insurance Institute for Highway Safety (IIHS) has seen a 7 percent reduction in crashes for vehicles with a basic forward-collision warning system, and a 14 to 15 percent reduction for those with automatic braking. ⁴ Google sibling company Waymo announced Wednesday a $2.5 billion investment round, which will go toward advancing its autonomous driving technology and growing its team. ⁵ Excluding robo-taxis, the share of new private vehicles with Level 2 entry systems will reach 47 percent by 2025 ⁶ The Global ADAS Market size is projected to grow from USD 27.2 billion in 2021 to USD 74.9 billion by 2030, at a CAGR of 11.9%. ⁷ Geofencing is used to define spatial boundaries, and detailed maps of the terrain and objects within the geofence are developed. The self-driving car projects the sensor data on top of the map to gather information and determine the safest path.}