Can you see me now? Blind spot estimation for autonomous vehicles using scenario-based simulation with random reference sensors

TL;DR

CanYouSeeMeNow introduces a Monte Carlo reference-sensor framework atop high-fidelity simulations to estimate sensor blind spots for autonomous vehicles. By unifying camera-depth and LiDAR data into 3D point clouds and probing coverage with randomly placed reference sensors, the method computes a nearest-neighbor blind-spot radius metric that can be visualized and aggregated across regions of interest. The approach supports evaluating sensor placement, mounting positions, and sensor resolutions under realistic geometry and occlusion, bridging simple geometric approximations and full SIL validation. It demonstrates actionable insights across camera and LiDAR configurations and integrates easily into scenario-based validation workflows.

Abstract

In this paper, we introduce a method for estimating blind spots for sensor setups of autonomous or automated vehicles and/or robotics applications. In comparison to previous methods that rely on geometric approximations, our presented approach provides more realistic coverage estimates by utilizing accurate and detailed 3D simulation environments. Our method leverages point clouds from LiDAR sensors or camera depth images from high-fidelity simulations of target scenarios to provide accurate and actionable visibility estimates. A Monte Carlo-based reference sensor simulation enables us to accurately estimate blind spot size as a metric of coverage, as well as detection probabilities of objects at arbitrary positions.

Figures (7)

• Figure 1: A camera setup with one front and two rear view cameras is simulated. The resulting coverage map \ref{['fig:page1:b']} provides an accurate estimate of the vehicle's field of view.
• Figure 2: Comparing our method to common methods of illustrating sensor setup coverage. The underlying camera setup is the same as illustrated in \ref{['fig:page1']}.
• Figure 3: An overview of the presented method: Sensor data from either camera or LiDAR sensors is compared to a reference pointcloud to estimate and illustrated the coverage of different sensor setups.
• Figure 4: Illustrating the need for a reference sensor: Without a reference sensor (top), a lack of sensor detections at a location may either be the result of lacking coverage (center) or missing target geometry (right). Probing the target geometry with a reference sensor (bottom) alleviates this issue.
• Figure 5: Computation of the blind spot radius $r_1$ and $r_2$ at the probe points $\boldsymbol{q}_1$ and $\boldsymbol{q}_2$.