Compressive sensing

Compressive sensing (CS), or compressed sampling, is a signal processing theory developed around 2006 by researchers like David Donoho and Emmanuel Candès. It posits that if a signal is sparse in a certain domain (e.g., Fourier or wavelet), it can be accurately reconstructed from far fewer samples than the Nyquist rate requires. This is achieved by taking non-adaptive linear measurements and then using optimization algorithms like L1-minimization for reconstruction. Within a risk management framework, CS is not a standard itself but an enabling technology. For instance, under ISO 21448 (SOTIF), sensor performance limitations are a primary source of risk. CS mitigates this by maintaining or enhancing perceptual capabilities with less data, directly addressing risks from processing delays or bandwidth constraints in ADAS and autonomous systems.

In automotive risk management, CS is applied to sensor data processing in ADAS to comply with ISO 26262 and ISO 21448. Implementation steps include: 1. **Risk Identification**: Identify hazardous scenarios arising from sensor data overload (e.g., from high-resolution LiDAR) causing ECU processing delays, as per ISO 21448 analysis. 2. **System Design**: Integrate CS algorithms into the sensor front-end or a dedicated processor (FPGA) to perform 'sampling and compression' simultaneously, transmitting only essential information. 3. **Reconstruction & Validation**: Deploy efficient reconstruction algorithms on the ECU and validate that the system's perception accuracy and reaction time meet the required ASIL under its Operational Design Domain (ODD). A Tier 1 supplier used CS to reduce LiDAR data transmission by 75%, cutting ECU load by 40% and improving system reaction time, thus significantly enhancing its SOTIF safety case.

Taiwanese enterprises face three key challenges in adopting CS for automotive use: 1. **Talent Gap in Algorithm-Hardware Co-design**: CS requires deep expertise in both complex mathematics and hardware integration (ASIC/FPGA), a skill set that is scarce in Taiwan's traditionally hardware-focused industry. 2. **Lack of Standardized Safety Validation**: Since standards like ISO 26262 do not explicitly define validation methods for CS, companies must develop their own safety cases, creating certification uncertainty. 3. **Immature Supply Chain**: The ecosystem for automotive-grade CS chips and development tools is limited and dominated by foreign firms, leading to high costs. To overcome this, firms should initiate university collaborations for talent, engage certification bodies early to co-develop validation plans, and form alliances with local IC designers to build a domestic supply chain.

Winners Consulting specializes in Compressive sensing for Taiwan enterprises, delivering compliant management systems within 90 days. Free consultation: https://winners.com.tw/contact