Behavioral Cloning Models Reality Check for Autonomous Driving

TL;DR

The real-world validation of state-of-the-art perception systems that utilize Behavior Cloning for lateral control, processing raw image data to predict steering commands and demonstrates promising potential for real-world applications.

Abstract

How effective are recent advancements in autonomous vehicle perception systems when applied to real-world autonomous vehicle control? While numerous vision-based autonomous vehicle systems have been trained and evaluated in simulated environments, there is a notable lack of real-world validation for these systems. This paper addresses this gap by presenting the real-world validation of state-of-the-art perception systems that utilize Behavior Cloning (BC) for lateral control, processing raw image data to predict steering commands. The dataset was collected using a scaled research vehicle and tested on various track setups. Experimental results demonstrate that these methods predict steering angles with low error margins in real-time, indicating promising potential for real-world applications.

Figures (15)

• Figure 1: CNN architecture
• Figure 2: Autoencoder network with encoded-decoded image result 8000 images with 80 epoch.After constructed image successfully, encoder network as in red highlighted part is used for BC
• Figure 3: Autobc network:encoder network combined with dense layer to generate behaviour cloning network
• Figure 4: AutoBC with Spatial Attention architecture. The RGB image is passed through a dual-head autoencoder with a ResNet18 backbone, with the first head outputting the reconstruction of the original image and the second head outputting an attention mask that highlights the lane boundaries. The attention mask is then used to provide spatial attention to the original image features, to highlight the salient regions. These resulting features are sent to a dense layer to predict the steering angle.
• Figure 5: Samples from image reconstruction. Top row is the original image, second row is the masked image, third row is the reconstructed image and the last row shows the masking applied.