Perception modules are integral in many modern autonomous systems, but their accuracy can be subject to the vagaries of the environment. In this paper, we propose a learning-based approach that can automatically characterize the error of a perception module from data and use this for safe control. The proposed approach constructs an inverse perception contract (IPC) which generates a set that contains the ground-truth value that is being estimated by the perception module, with high probability. We apply the proposed approach to study a vision pipeline deployed on a quadcopter. With the proposed approach, we successfully constructed an IPC for the vision pipeline. We then designed a control algorithm that utilizes the learned IPC, with the goal of landing the quadcopter safely on a landing pad. Experiments show that with the learned IPC, the control algorithm safely landed the quadcopter despite the error from the perception module, while the baseline algorithm without using the learned IPC failed to do so.

Deep learning (DL) models are becoming effective in solving computer-vision tasks such as semantic segmentation, object tracking, and pose estimation on real-world captured images. Reliability analysis of autonomous systems that use these DL models as part of their perception systems have to account for the performance of these models. Autonomous systems with traditional sensors have tried-and-tested reliability assessment processes with modular design, unit tests, system integration, compositional verification, certification, etc. In contrast, DL perception modules relies on data-driven or learned models. These models do not capture uncertainty and often lack robustness. Also, these models are often updated throughout the lifecycle of the product when new data sets become available. However, the integration of an updated DL-based perception requires a reboot and start afresh of the reliability assessment and operation processes for autonomous systems. In this paper, we discuss three challenges related to specifying, verifying, and operating systems that incorporate DL-based perception. We illustrate these challenges through two concrete and open source examples.

Abstract We present , a tool that uses neural networks for predicting reachable sets from executions of a dynamical system. Unlike existing reachability tools, computes areachability functionthat outputs an accurate over-approximation of the reachable set foranyinitial set in a parameterized family. Such reachability functions are useful for online monitoring, verification, and safe planning. implements empirical risk minimization for learning reachability functions. We discuss the design rationale behind the optimization problem and establish that the computed output is probably approximately correct. Our experimental evaluations over a variety of systems show promise. can learn accurate reachability functions for complex nonlinear systems, including some that are beyond existing methods. From a learned reachability function, arbitrary reachtubes can be computed in milliseconds. is available at https://github.com/sundw2014/NeuReach .

Establishing correct correspondences between two images should consider both local and global spatial context. Given putative correspondences of feature points in two views, in this paper, we propose Order-Aware Network, which infers the probabilities of correspondences being inliers and regresses the relative pose encoded by the essential or fundamental matrix. Specifically, this proposed network is built hierarchically and comprises three operations. First, to capture the local context of sparse correspondences, the network clusters unordered input correspondences by learning a soft assignment matrix. These clusters are in canonical order and invariant to input permutations. Next, the clusters are spatially correlated to encode the global context of correspondences. After that, the context-encoded clusters are interpolated back to the original size and position to build a hierarchical architecture. We intensively experiment on both outdoor and indoor datasets. The accuracy of the two-view geometry and correspondences are significantly improved over the state-of-the-arts. Besides, based on the proposed method and advanced local feature, we won the first place in CVPR 2019 image matching workshop challenge and also achieve state-of-the-art results in the Visual Localization benchmark. Code is available at https://github.com/zjhthu/OANet.

Convolutional Neural Networks (CNN) for object detection, lane detection, and segmentation now sit at the head of most autonomy pipelines, and yet, their safety analysis remains an important challenge. Formal analysis of perception models is fundamentally difficult because their correctness is hard if not impossible to specify. We present a technique for inferring intelligible and safe abstractions for perception models from system-level safety requirements, data, and program analysis of the modules that are downstream from perception. The technique can help tradeoff safety, size, and precision, in creating abstractions and the subsequent verification. We apply the method to two significant case studies based on high-fidelity simulations (a) a vision-based lane keeping controller for an autonomous vehicle and (b) a controller for an agricultural robot. We show how the generated abstractions can be composed with the downstream modules and then the resulting abstract system can be verified using program analysis tools like CBMC. Detailed evaluations of the impacts of size, safety requirements, and the environmental parameters (e.g., lighting, road surface, plant type) on the precision of the generated abstractions suggest that the approach can help guide the search for corner cases and safe operating envelops.

Convolutional Neural Networks (CNNs) have become deeper and more complicated compared with the pioneering AlexNet. However, current prevailing training scheme follows the previous way of adding supervision to the last layer of the network only and propagating error information up layer-by-layer. In this paper, we propose Deeply-supervised Knowledge Synergy (DKS), a new method aiming to train CNNs with improved generalization ability for image classification tasks without introducing extra computational cost during inference. Inspired by the deeply-supervised learning scheme, we first append auxiliary supervision branches on top of certain intermediate network layers. While properly using auxiliary supervision can improve model accuracy to some degree, we go one step further to explore the possibility of utilizing the probabilistic knowledge dynamically learnt by the classifiers connected to the backbone network as a new regularization to improve the training. A novel synergy loss, which considers pairwise knowledge matching among all supervision branches, is presented. Intriguingly, it enables dense pairwise knowledge matching operations in both top-down and bottom-up directions at each training iteration, resembling a dynamic synergy process for the same task. We evaluate DKS on image classification datasets using state-of-the-art CNN architectures, and show that the models trained with it are consistently better than the corresponding counterparts. For instance, on the ImageNet classification benchmark, our ResNet-152 model outperforms the baseline model with a 1.47% margin in Top-1 accuracy. Code is available at https://github.com/sundw2014/DKS.

Establishing correspondences between two images requires both local and global spatial context. Given putative correspondences of feature points in two views, in this paper, we propose Order-Aware Network, which infers the probabilities of correspondences being inliers and regresses the relative pose encoded by the essential matrix. Specifically, this proposed network is built hierarchically and comprises three novel operations. First, to capture the local context of sparse correspondences, the network clusters unordered input correspondences by learning a soft assignment matrix. These clusters are in a canonical order and invariant to input permutations. Next, the clusters are spatially correlated to form the global context of correspondences. After that, the context-encoded clusters are recovered back to the original size through a proposed upsampling operator. We intensively experiment on both outdoor and indoor datasets. The accuracy of the two-view geometry and correspondences are significantly improved over the state-of-the-arts.

We propose a unified mathematical model for multilayer-multiframe compressive light field displays that supports both attenuation-based and polarization-based architectures. We show that the light field decomposition of such a display can be cast as a bound constrained nonlinear matrix optimization problem. Efficient light field decomposition algorithms are developed using the limited-memory BFGS (L-BFGS) method for automultiscopic displays with high resolution and high image fidelity. In addition, this framework is the first to support multilayer polarization-based compressive light field displays with time multiplexing. This new architecture significantly reduces artifacts compared with attenuation-based multilayer-multiframe displays; thus, it can allow the requirements regarding the number of layers or the refresh rate to be relaxed. We verify the proposed methods by constructing two 3-layer prototypes using high-speed LCDs, one based on the attenuation architecture and one based on the polarization architecture. Moreover, an efficient CUDA-based program is implemented. Our displays can produce images with higher spatial resolution with thinner form factors compared with traditional automultiscopic displays in both simulations and experiments.