What Do They See? Interpreting Complex Road Scenarios Through the Eyes of Vision-Language-Action Models for Safe and Trustworthy Autonomous Vehicle Learning
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine LearningRobotics
Summary
The authors studied how self-driving car models decide what to do by removing objects like cars or pedestrians from the camera images and seeing how the model's driving plans change. They used a special technique to erase these objects realistically and tested it on a popular driving prediction model across many scenes. They found that cars and people near the car’s planned path greatly influence decisions, and traffic lights also have a strong effect despite being small in the image. Interestingly, sometimes the model responded to things that humans might ignore, raising questions about how the model internally processes visual information. The authors also looked inside the model's layers to better understand how these objects affect its decision-making, aiming to make self-driving systems more understandable and trustworthy.
Authors
Kalpana Panda, Wesley Maia, Vinti Agarwal, Ross Greer
Abstract
End-to-end autonomous driving models are now able to navigate complex road scenarios, mapping raw sensor observations directly to observed paths for open-loop evaluation and often effective driving in closed-loop evaluation. Yet the internal logic of these safety-critical systems remains largely opaque, due to the complexity of traffic scenes. We propose a counterfactual ablation framework called Counterfactual Vision Action Analysis (CVAA) that systematically removes individual detected objects from front-camera images using photorealistic generative inpainting to prepare counterfactual sets to evaluate the difference in the model's response. This isolates the causal effect of each object's presence on the model's planning behaviour. Applied to the Alpamayo 1 trajectory predictor across 210 nuScenes driving scenes, we create a dataset Counter -nuScenes, using which we see that vehicles and pedestrians within the model's 'path' dominate causal influence as expected, while traffic lights, as expected, exert disproportionate effect relative to their image footprint. However, we also find cases where the model responds strongly to objects a human driver would consider irrelevant. This brings forth a deeper question: does the model itself view the scene as a sum of individual objects influencing the outcome, or does it encode an entirely different set of internal features that do not correspond to human-legible scene elements? To further understand this, we compare intermediate representations of original and inpainted image pairs using mechanistic interpretability techniques and examine the effect of the removal through the various model layers. Together, these two stages offer a path from behavioral auditing to representational understanding, creating explainable driving systems and solidifying human-AI trust.