Non-Volatile Memory for ADAS: Protecting Critical Data in Pre-Crash Moments
This article explains the importance of data recording in supporting ADAS safety and countermeasures in the event of power loss. It outlines design requirements for reliably protecting data just before collision and summarizes the challenges of conventional memory. The advantages of FeRAM, with its high-speed writes, are also introduced.
Safety Control and the Role of Data in ADAS
ADAS is a system that integrates sensor information to achieve accident avoidance by making split-second decisions and executing control actions. Within this system, data is not merely input; it plays a crucial role as the basis for ensuring safety. Especially in recent years, from the perspective of accountability and compliance with legal regulations, the handling of data, including the decision-making process, has become an indispensable design consideration.
Sensor Fusion and Real-Time Decision-Making
ADAS integrates information obtained from multiple sensors, such as cameras, millimeter-wave radar, and LiDAR, to recognize the surrounding environment with high accuracy. This sensor fusion enables situational assessment that would be difficult using a single sensor, reducing false positives and improving safety. In addition, these processes must be completed within milliseconds, making real-time performance critical. Because the acquired data is immediately reflected in decision-making, both high-speed processing and high reliability are required.
Decision-Making Process in ADAS
ADAS decision-making proceeds through stages of situational awareness, risk assessment, and control decisions based on input sensor data. For example, when an obstacle is detected, the system computes time-to-collision (TTC) from object tracking and determines the timing of interventions such as automatic emergency braking. Because this series of processes is executed automatically based on algorithms, the accuracy of the data that forms the basis of these decisions is extremely important. Furthermore, the decision results must be verifiable later, so systems must be architected for data recording from the outset.
Why Are Operation Logs Necessary?
In ADAS, operation logs are not merely records, but a crucial element for proving safety. In the event of a malfunction or accident, it is necessary to be able to trace back and review sensor inputs, decision processes, and control actions to identify the cause. Especially with international regulatory mandates for event data recorders (EDR) and data storage systems for automated driving (DSSAD), vehicle behavior must be recorded in tamper-resistant form. Furthermore, since it is used for algorithm improvements during the development stage and analysis of field data, a highly reliable recording mechanism is required.
Data Recording Requirements for ADAS
ADAS requires data recording to be performed in parallel with real-time control, and these requirements are more stringent than those of conventional automotive systems. In particular, to reliably preserve the system state immediately before a collision, a mechanism capable of completing write operations even during momentary power interruptions is essential. Furthermore, because the recorded data is used for subsequent analysis and safety verification, it must provide a high level of data integrity and reliability.
The Importance of Pre-Collision Data
Data from the few seconds preceding a detected collision is extremely important for identifying the cause of an accident and verifying system operation. This data includes vehicle speed, yaw rate, brake status, and the results of surrounding object recognition, providing the basis for evaluating whether the ADAS made appropriate decisions. In addition, the importance of data retention has increased in recent years from the perspective of regulatory compliance, making a reliable data recording mechanism essential. Therefore, the design must guarantee recording without delay or loss.
Fail-Safe Design and Log Storage
Log storage in ADAS is closely related to fail-safe design. Even when a system malfunction occurs, the system must transition to a safe state while capturing a record of the fault condition. In particular, during events such as power loss or system resets, data being written may be lost when conventional volatile memory is used. Therefore, a mechanism capable of reliably completing write operations within a short time using only the energy stored in backup capacitors is essential, requiring a system architecture that incorporates log storage from the design stage.
Utilization for AI and Algorithm Verification
Since ADAS uses AI and advanced algorithms, data for verifying their operation is essential. Collected logs are used for analyzing false positives and improving performance, contributing to increased system reliability. Data obtained in real-world vehicle environments is particularly valuable because it is difficult to reproduce in simulations. Furthermore, logs play a crucial role in explaining the rationale behind AI decisions. Since this recorded data is utilized in both development and operation, a robust, high-integrity storage foundation is essential.
Non-Volatile Memory Technology Supporting ADAS
Non-volatile memory characteristics play a vital role in enabling data recording in ADAS. In particular, the ability to retain data even during power loss, along with the ability to reliably complete write operations in a short time, is required. Cases that conventional memories cannot adequately address are increasing, making the selection of appropriate memory for each application a critical factor in ensuring safety.
Maintaining Records Through Power Loss
In ADAS, power can be lost without warning in events such as collisions, requiring a mechanism to reliably record data immediately preceding such an event. Therefore, the system is designed to immediately execute a write operation using the charge stored in a capacitor upon detecting a power loss. However, writing takes time with conventional memory, so high-speed memory is required to complete the write within the few milliseconds available before the
supply voltage falls below the operating threshold. Non-volatile memory, in particular, with its low write latency, is an effective means of reducing the risk of data loss.
Challenges and Limitations of Conventional Memory
While flash memory and EEPROM are widely used, they present challenges for ADAS applications in terms of write speed and endurance. Flash memory, in particular, requires pre-write data erasure, taking several milliseconds or more to complete. To compensate for this delay, a large backup capacitor must be mounted on the board, increasing board area and cost. Furthermore, the limited program/erase (P/E) cycle count makes it unsuitable for continuous recording where the latest data must be constantly overwritten. These characteristics are constraints for applications requiring data storage just before a collision, increasing the need for more reliable memory technology.
Why FeRAM Is Suitable for ADAS
FeRAM is a non-volatile memory technology that combines high-speed write performance with high endurance, making it well suited to ADAS applications. Because it does not require an erase operation before writing and completes writes in mere hundreds of nanoseconds—orders of magnitude faster than flash—it reliably captures data in the brief window before power is lost while dramatically shrinking the required backup capacitance. In addition, it offers virtually unlimited write endurance, making it suitable for frequent log recording. Furthermore, its low power consumption makes it well suited for use with backup power supplies. These characteristics make FeRAM an effective choice as a highly reliable data recording technology.
Non-Volatile Memory Design Essentials for ADAS Reliability
ADAS safety depends not only on control algorithms but also on the data recording infrastructure that supports them. In particular, the reliability of data recording during abnormal events is a critical factor affecting the overall reliability of the system. Therefore, non-volatile memory should be regarded not merely as a storage medium but as an integral part of the system’s safety design.
The Relationship Between ADAS and Data Recording
While ADAS performs real-time environmental perception and vehicle control, it also incorporates data recording functions that enable its operation to be verified afterward. These functions include not only routine logs but also mechanisms for preserving critical data before and after a collision. Event-specific data recording is particularly essential for accident analysis and regulatory compliance. In other words, data recording in ADAS is not merely an auxiliary function but a foundation that supports both safety and accountability, and it should be considered as an integral part of the system architecture from the earliest stages of system design.
Key Considerations for Memory Selection
When selecting non-volatile memory for ADAS applications, write reliability and real-time performance are more important than capacity or cost. The decisive metric is whether all critical data can be committed within the window before power is lost, and memory with write delays carries the risk of requiring large capacitors. Furthermore, high write endurance is required to handle frequent log updates. Additionally, compatibility with the entire system must be considered, requiring a design that integrates with backup power supplies and control logic. Therefore, the memory should be evaluated not in isolation but as part of the overall architecture.
Future ADAS Evolution and Memory Technology
ADAS will continue to evolve alongside advances in autonomous driving technology, resulting in dramatic increases in both the volume of data processed and the frequency of processing. Consequently, the amount of information that must be recorded will also increase, creating a need for higher-performance non-volatile memory than ever before. In particular, more detailed logging will be required to enable the AI decision-making process to be traced, further increasing the importance of data recording. Against this backdrop, non-volatile memory is positioned not merely as an auxiliary component but as a foundational technology for the evolution of ADAS—one where continued innovation will be essential.
RAMXEED AEC-Q100-Qualified SPI FeRAM Family (1–4 Mbit)
RAMXEED AEC-Q100-Qualified SPI FeRAM Family (64–512 Kbit)
