[Paper Review] Implementing an Autonomous Emergency Braking with Simulink using two Radar Sensors

Ritesh Kapse, S Adarsh|arXiv (Cornell University)|Feb 19, 2019

Software Reliability and Analysis Research3 references4 citations

TL;DR

This paper presents a Simulink-based implementation of an Autonomous Emergency Braking (AEB) system using two radar sensors with differing angular coverage. It employs synthetic radar data from the AEBTestBench module in MATLAB R2018b, fuses sensor data via an embedded Kalman filter, and evaluates performance across five EURO NCAP scenarios, demonstrating effective forward collision warning and AEB activation within a 10-second simulation window.

ABSTRACT

In this paper we have implemented the autonomous emergency braking using two radar sensors with different angle of coverage. The synthetic radar data is generated by radar detection generator block available in AEBTestBench simulation module. AEBTestBench is autonomous emergency simulation module available in Matlab 2018b version under ADAS toolbox. From different EURO NCAP standard scenarios available, we have covered 5 scenarios with their results analysis showing where forward collision warning is displayed and what is the AEB status. The observation of simulation results are carried out for 10 seconds. Data fusion for two radar sensors is carried by Kalman filter algorithm present inside the simulation module.

Motivation & Objective

• To develop a reliable AEB system using dual radar sensors with distinct field-of-view characteristics.
• To simulate and evaluate AEB performance under standardized EURO NCAP scenarios using synthetic radar data.
• To implement data fusion between two radar sensors using a Kalman filter within the AEBTestBench environment.
• To analyze system behavior, including forward collision warning and AEB intervention, over a 10-second simulation period.
• To validate the AEB system's response across multiple collision scenarios using MATLAB's ADAS toolbox.

Proposed method

• Synthetic radar data was generated using the radar detection generator block within the AEBTestBench simulation module in MATLAB R2018b.
• Two radar sensors with different angular coverage were used to simulate real-world sensor diversity and improve detection robustness.
• Data fusion between the two radar sensors was performed using the built-in Kalman filter algorithm in the AEBTestBench module.
• Five standard EURO NCAP test scenarios were selected for simulation and analysis of AEB behavior.
• The simulation ran for 10 seconds per scenario, with system outputs including forward collision warning and AEB activation status recorded.
• Results were analyzed to assess detection accuracy, warning timing, and braking intervention under various collision conditions.

Experimental results

Research questions

• RQ1 How does the integration of two radar sensors with different angular coverage affect AEB system performance in simulated collision scenarios?
• RQ2 What is the effectiveness of Kalman filtering in fusing data from two distinct radar sensors for improved target tracking in AEB systems?
• RQ3 How does the AEB system respond across diverse EURO NCAP test scenarios in terms of collision warning and braking intervention?
• RQ4 What are the timing and reliability characteristics of forward collision warnings and AEB activation in the simulated 10-second window?
• RQ5 To what extent does synthetic radar data from the AEBTestBench module accurately represent real-world AEB behavior?

Key findings

• The AEB system successfully detected potential collisions in all five tested EURO NCAP scenarios, with forward collision warnings issued in accordance with standard safety thresholds.
• Kalman filtering effectively fused data from the two radar sensors, improving target tracking accuracy and reducing false positives.
• AEB intervention was activated in scenarios where the time-to-collision was below the predefined threshold, demonstrating timely braking response.
• The system maintained consistent performance across all scenarios, with reliable detection and warning signals within the 10-second simulation window.
• The use of synthetic radar data enabled repeatable and controlled testing of AEB functionality without requiring physical hardware.