ADAS Intel Weekly: Comprehensive ADAS Radar Intelligence

# ADAS Intel Weekly: Comprehensive ADAS Radar Intelligence **Date:** July 20, 2026 **Focus:** 4D Imaging Radar, Sensor Fusion, and Regulatory Compliance --- ## 1. Vendor Developments & Innovations ### 🚀 Altos Radar - **Strategic Move:** Appointed former Bosch radar lead, Dr. Mingkang Li, as President to accelerate 4D imaging radar technology and European expansion. - **Technical Edge:** Utilizing an **ASIC-based approach** to fast-track viable 4D systems. - **Product Highlight:** The **Altos V4 (16Tx16Rx)** 4D main radar focuses on ultra-high point cloud imaging and all-weather robustness (heavy rain, snow, fog), filling the gap left by cameras. ### 🛡️ Valeo - **Breakthrough Program:** Secured a major program for **unsupervised Highway Pilot at 130kph** (SAE Level 3). - **System Integration:** The system uniquely combines high-resolution LiDAR, cameras, and a proprietary radar system to enable driving without a lead vehicle. - **Timeline:** Production is slated to begin in **2028**. ### 🏢 Industry Landscape (Bosch, Aptiv, ZF) - The industry is transitioning into the **"Compute Era"** of ADAS, where radar is evolving from simple detection tools to rich perception sensors providing dense point clouds. - The 4D Imaging Radar market is projected to grow from ~$4.3B in 2026 to **$13.6B by 2033** (CAGR ~21.5%). --- ## 2. Sensor Fusion & Academic-Industry Collaborations ### 🎓 SJTU & CSIF-ISIF Initiatives - **The FOOD Platform:** The **Fusion Oriented Open-Access Data (FOOD) platform**, hosted at [cadar.ai](https://cadar.ai), has emerged as a critical hub for standardization and benchmarking of fusion algorithms for autonomous vehicles. - **Collaborative Framework:** SJTU has established a special committee for "Information Fusion for Autonomous Systems," serving as a bridge between academic research and industry-standard benchmarks. - **Infrastructure:** Development of a "Smart Data Center" to provide high-performance storage and synchronized playback of multi-modal data streams. --- ## 3. Technical Deep Dive: Radar Fusion ### 🛠️ Fusion Algorithms & Processing - **Distributed Fusion:** Recent research highlights **multi-radar distributed fusion** aided by multi-feature information to improve target detection accuracy. - **Algorithm Stack:** Implementation of **Bayesian, Dempster-Shafer, and Averaging** algorithms for data fusion. - **Preprocessing:** Heavy emphasis on **outlier detection** to eliminate erroneous data points before the fusion stage to prevent "ghost" targets. - **Key Challenges:** Addressing sensor communication latency, extended object data association, and fusion with incomplete information. --- ## 4. Open-Access Data Sources - **FOOD Platform ([cadar.ai](https://cadar.ai)):** Identified as a primary potential source for validating fusion algorithms. It provides a visualizer and a standardized data pipeline (Board Collection $\rightarrow$ Server Storage $\rightarrow$ Frontend Visualization). --- ## 5. Regulatory Analysis: EU AI Act - **Effective Date:** Most provisions become legally applicable on **August 2, 2026**. - **Risk Classification:** AI systems used as safety components in AVs are likely to be classified as **"High-Risk AI Systems."** - **Requirements:** Strict mandates on **data governance, risk management, and human oversight**. - **Liability Impact:** A shift toward increased accountability for manufacturers and deployers. However, the Act allows for **regulatory sandboxes** to enable innovation without immediate full-scale compliance burdens during the testing phase. --- ## 6. Summary of Industry Trends | Trend | Shift | Impact | | :--- | :--- | :--- | | **Perception** | Detection $\rightarrow$ Imaging | Radar now provides "vision-like" point clouds, reducing reliance on LiDAR for some L2+/L3 tasks. | | **Architecture** | Hardware-Centric $\rightarrow$ Software-Defined | Shift toward compute-heavy ADAS where software defines the radar's capability. | | **Fusion** | Late Fusion $\rightarrow$ Hybrid/Deep Fusion | Integration of raw data streams (multi-feature) rather than just object lists. | | **Compliance** | Self-Regulation $\rightarrow$ Strict AI Governance | EU AI Act forces transparency and rigorous validation of "black box" AI models. |

# ADAS Intel Weekly: Comprehensive ADAS Radar Intelligence
**Date:** July 20, 2026
**Focus:** 4D Imaging Radar, Sensor Fusion, and Regulatory Compliance

---

## 1. Vendor Developments & Innovations

### 🚀 Altos Radar
- **Strategic Move:** Appointed former Bosch radar lead, Dr. Mingkang Li, as President to accelerate 4D imaging radar technology and European expansion.
- **Technical Edge:** Utilizing an **ASIC-based approach** to fast-track viable 4D systems.
- **Product Highlight:** The **Altos V4 (16Tx16Rx)** 4D main radar focuses on ultra-high point cloud imaging and all-weather robustness (heavy rain, snow, fog), filling the gap left by cameras.

### 🛡️ Valeo
- **Breakthrough Program:** Secured a major program for **unsupervised Highway Pilot at 130kph** (SAE Level 3). 
- **System Integration:** The system uniquely combines high-resolution LiDAR, cameras, and a proprietary radar system to enable driving without a lead vehicle.
- **Timeline:** Production is slated to begin in **2028**.

### 🏢 Industry Landscape (Bosch, Aptiv, ZF)
- The industry is transitioning into the **"Compute Era"** of ADAS, where radar is evolving from simple detection tools to rich perception sensors providing dense point clouds.
- The 4D Imaging Radar market is projected to grow from ~$4.3B in 2026 to **$13.6B by 2033** (CAGR ~21.5%).

---

## 2. Sensor Fusion & Academic-Industry Collaborations

### 🎓 SJTU & CSIF-ISIF Initiatives
- **The FOOD Platform:** The **Fusion Oriented Open-Access Data (FOOD) platform**, hosted at [cadar.ai](https://cadar.ai), has emerged as a critical hub for standardization and benchmarking of fusion algorithms for autonomous vehicles.
- **Collaborative Framework:** SJTU has established a special committee for "Information Fusion for Autonomous Systems," serving as a bridge between academic research and industry-standard benchmarks.
- **Infrastructure:** Development of a "Smart Data Center" to provide high-performance storage and synchronized playback of multi-modal data streams.

---

## 3. Technical Deep Dive: Radar Fusion

### 🛠️ Fusion Algorithms & Processing
- **Distributed Fusion:** Recent research highlights **multi-radar distributed fusion** aided by multi-feature information to improve target detection accuracy.
- **Algorithm Stack:** Implementation of **Bayesian, Dempster-Shafer, and Averaging** algorithms for data fusion.
- **Preprocessing:** Heavy emphasis on **outlier detection** to eliminate erroneous data points before the fusion stage to prevent "ghost" targets.
- **Key Challenges:** Addressing sensor communication latency, extended object data association, and fusion with incomplete information.

---

## 4. Open-Access Data Sources

- **FOOD Platform ([cadar.ai](https://cadar.ai)):** Identified as a primary potential source for validating fusion algorithms. It provides a visualizer and a standardized data pipeline (Board Collection $\rightarrow$ Server Storage $\rightarrow$ Frontend Visualization).

---

## 5. Regulatory Analysis: EU AI Act

- **Effective Date:** Most provisions become legally applicable on **August 2, 2026**.
- **Risk Classification:** AI systems used as safety components in AVs are likely to be classified as **"High-Risk AI Systems."**
- **Requirements:** Strict mandates on **data governance, risk management, and human oversight**.
- **Liability Impact:** A shift toward increased accountability for manufacturers and deployers. However, the Act allows for **regulatory sandboxes** to enable innovation without immediate full-scale compliance burdens during the testing phase.

---

## 6. Summary of Industry Trends

| Trend | Shift | Impact |
| :--- | :--- | :--- |
| **Perception** | Detection $\rightarrow$ Imaging | Radar now provides "vision-like" point clouds, reducing reliance on LiDAR for some L2+/L3 tasks. |
| **Architecture** | Hardware-Centric $\rightarrow$ Software-Defined | Shift toward compute-heavy ADAS where software defines the radar's capability. |
| **Fusion** | Late Fusion $\rightarrow$ Hybrid/Deep Fusion | Integration of raw data streams (multi-feature) rather than just object lists. |
| **Compliance** | Self-Regulation $\rightarrow$ Strict AI Governance | EU AI Act forces transparency and rigorous validation of "black box" AI models. |

Comments