Learning Vision-Based Neural Network Controllers with Semi-Probabilistic Safety Guarantees

2025-02-28

AI Summary

To address the lack of formal safety guarantees for autonomous systems operating on visual inputs, this paper proposes the first semi-probabilistic safety verification framework integrating reachability analysis, conditional generative adversarial networks (cGANs), and distribution-free tail-bound estimation. We further design an end-to-end training paradigm synergizing a safety-aware loss function, critical-sample active sampling, and curriculum learning. The method jointly achieves high-fidelity perception modeling in high-dimensional visual spaces and rigorous, verifiable safety. Evaluated on X-Plane 11 landing, CARLA lane-following, and F1Tenth physical-platform tasks, it delivers semi-probabilistic safety guarantees with ≥99.9% confidence while matching the nominal performance of unconstrained models. Key contributions include: (i) the first scalable semi-probabilistic verification architecture; (ii) a perception–safety co-optimization paradigm; and (iii) empirical validation of high-confidence safety bounds under real-hardware closed-loop operation.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Safety and Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Ensuring safety in autonomous systems with vision-based control remains a critical challenge due to the high dimensionality of image inputs and the fact that the relationship between true system state and its visual manifestation is unknown. Existing methods for learning-based control in such settings typically lack formal safety guarantees. To address this challenge, we introduce a novel semi-probabilistic verification framework that integrates reachability analysis with conditional generative adversarial networks and distribution-free tail bounds to enable efficient and scalable verification of vision-based neural network controllers. Next, we develop a gradient-based training approach that employs a novel safety loss function, safety-aware data-sampling strategy to efficiently select and store critical training examples, and curriculum learning, to efficiently synthesize safe controllers in the semi-probabilistic framework. Empirical evaluations in X-Plane 11 airplane landing simulation, CARLA-simulated autonomous lane following, and F1Tenth lane following in a physical visually-rich miniature environment demonstrate the effectiveness of our method in achieving formal safety guarantees while maintaining strong nominal performance. Our code is available at https://github.com/xhOwenMa/SPVT.

Problem

• Ensuring safety in vision-based autonomous systems
• Lack of formal safety guarantees in existing methods
• High dimensionality and unknown state-visual relationship challenges

Innovation

• Semi-probabilistic verification framework integration
• Gradient-based training with safety loss function
• Safety-aware data-sampling and curriculum learning