Car Drag Coefficient Prediction from 3D Point Clouds Using a Slice-Based Surrogate Model

TL;DR

This work tackles the need for rapid and accurate prediction of the drag coefficient $C_d$ from 3D vehicle geometry, addressing the bottlenecks of CFD and wind-tunnel testing. It introduces a slice-based surrogate that converts a 3D point cloud into an ordered sequence of 2D cross-sections along the streamwise axis, encodes each slice with a lightweight PointNet2D, and models the longitudinal geometric evolution with a Bi-LSTM before regressing $C_d$ with an MLP. On the DrivAerNet++ dataset, the method achieves $R^2=0.9528$ and MAE $=6.046 imes10^{-3}$ with an inference time of ~0.025 s per sample on a consumer-grade GPU, using about 2.79M parameters. This approach offers a transparent, efficient alternative to heavier 3D CNNs or transformer-based point-cloud models, enabling fast design iteration and interpretable attribution of drag along vehicle length.

Abstract

The automotive industry's pursuit of enhanced fuel economy and performance necessitates efficient aerodynamic design. However, traditional evaluation methods such as computational fluid dynamics (CFD) and wind tunnel testing are resource intensive, hindering rapid iteration in the early design stages. Machine learning-based surrogate models offer a promising alternative, yet many existing approaches suffer from high computational complexity, limited interpretability, or insufficient accuracy for detailed geometric inputs. This paper introduces a novel lightweight surrogate model for the prediction of the aerodynamic drag coefficient (Cd) based on a sequential slice-wise processing of the geometry of the 3D vehicle. Inspired by medical imaging, 3D point clouds of vehicles are decomposed into an ordered sequence of 2D cross-sectional slices along the stream-wise axis. Each slice is encoded by a lightweight PointNet2D module, and the sequence of slice embeddings is processed by a bidirectional LSTM to capture longitudinal geometric evolution. The model, trained and evaluated on the DrivAerNet++ dataset, achieves a high coefficient of determination (R^2 > 0.9528) and a low mean absolute error (MAE approx 6.046 x 10^{-3}) in Cd prediction. With an inference time of approximately 0.025 seconds per sample on a consumer-grade GPU, our approach provides fast, accurate, and interpretable aerodynamic feedback, facilitating more agile and informed automotive design exploration.

Figures (5)

• Figure 1: Illustration of 80 streamwise (x-axis) slices extracted from a car’s point cloud. This isometric side view demonstrates how dense slicing captures detailed shape variation from front to back.
• Figure 2: Visualization of 80 streamwise (X-axis) cross-sectional slices extracted from a vehicle's point cloud. This grid displays the individual 2D point sets, sequentially arranged to highlight the progression of the vehicle's contour from front to back, as used for feature extraction.
• Figure 3: Complete model architecture of the proposed sequential slice-based drag prediction model. 3D point clouds are sliced. Each slice is encoded by PointNet2D. The sequence of embeddings is processed by a Bi-LSTM and an MLP regresses $C_d$.
• Figure 4: Training dynamics: (a) Training loss curve. (b) Validation $R^2$ score over epochs. Best validation $R^2$ was achieved at epoch 68.
• Figure 5: Test set performance analysis: (a) Scatter plot showing strong correlation between predicted and true $C_d$ values. (b) Histogram of prediction errors, centered near zero with small spread.