Deep learning has emerged as a promising paradigm to give access to highly accurate predictions of molecular and material properties. A common short-coming shared by current approaches, however, is that neural networks only give point estimates of their predictions and do not come with predictive uncertainties associated with these estimates. Existing uncertainty quantification efforts have primarily leveraged the standard deviation of predictions across an ensemble of independently trained neural networks. This incurs a large computational overhead in both training and prediction, resulting in order-of-magnitude more expensive predictions. Here, we propose a method to estimate the predictive uncertainty based on a single neural network without the need for an ensemble. This allows us to obtain uncertainty estimates with virtually no additional computational overhead over standard training and inference. We demonstrate that the quality of the uncertainty estimates matches those obtained from deep ensembles. We further examine the uncertainty estimates of our methods and deep ensembles across the configuration space of our test system and compare the uncertainties to the potential energy surface. Finally, we study the efficacy of the method in an active learning setting and find the results to match an ensemble-based strategy at order-of-magnitude reduced computational cost.
This content will become publicly available on August 16, 2024
Robots are active agents that operate in dynamic scenarios with noisy sensors. Predictions based on these noisy sensor measurements often lead to errors and can be unreliable. To this end, roboticists have used fusion methods using multiple observations. Lately, neural networks have dominated the accuracy charts for perception-driven predictions for robotic decision-making and often lack uncertainty metrics associated with the predictions. Here, we present a mathematical formulation to obtain the heteroscedastic aleatoric uncertainty of any arbitrary distribution without prior knowledge about the data. The approach has no prior assumptions about the prediction labels and is agnostic to network architecture. Furthermore, our class of networks, Ajna, adds minimal computation and requires only a small change to the loss function while training neural networks to obtain uncertainty of predictions, enabling real-time operation even on resource-constrained robots. In addition, we study the informational cues present in the uncertainties of predicted values and their utility in the unification of common robotics problems. In particular, we present an approach to dodge dynamic obstacles, navigate through a cluttered scene, fly through unknown gaps, and segment an object pile, without computing depth but rather using the uncertainties of optical flow obtained from a monocular camera with onboard sensing and computation. We successfully evaluate and demonstrate the proposed Ajna network on four aforementioned common robotics and computer vision tasks and show comparable results to methods directly using depth. Our work demonstrates a generalized deep uncertainty method and demonstrates its utilization in robotics applications.
more » « less- Award ID(s):
- 2020624
- NSF-PAR ID:
- 10484646
- Publisher / Repository:
- American Association for the Advancement of Science
- Date Published:
- Journal Name:
- Science Robotics
- Volume:
- 8
- Issue:
- 81
- ISSN:
- 2470-9476
- Page Range / eLocation ID:
- eadd5139
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
-
null (Ed.)Traditional deep neural networks (NNs) have significantly contributed to the state-of-the-art performance in the task of classification under various application domains. However, NNs have not considered inherent uncertainty in data associated with the class probabilities where misclassification under uncertainty may easily introduce high risk in decision making in real-world contexts (e.g., misclassification of objects in roads leads to serious accidents). Unlike Bayesian NN that indirectly infer uncertainty through weight uncertainties, evidential NNs (ENNs) have been recently proposed to explicitly model the uncertainty of class probabilities and use them for classification tasks. An ENN offers the formulation of the predictions of NNs as subjective opinions and learns the function by collecting an amount of evidence that can form the subjective opinions by a deterministic NN from data. However, the ENN is trained as a black box without explicitly considering inherent uncertainty in data with their different root causes, such as vacuity (i.e., uncertainty due to a lack of evidence) or dissonance (i.e., uncertainty due to conflicting evidence). By considering the multidimensional uncertainty, we proposed a novel uncertainty-aware evidential NN called WGAN-ENN (WENN) for solving an out-of-distribution (OOD) detection problem. We took a hybrid approach that combines Wasserstein Generative Adversarial Network (WGAN) with ENNs to jointly train a model with prior knowledge of a certain class, which has high vacuity for OOD samples. Via extensive empirical experiments based on both synthetic and real-world datasets, we demonstrated that the estimation of uncertainty by WENN can significantly help distinguish OOD samples from boundary samples. WENN outperformed in OOD detection when compared with other competitive counterpartsmore » « less
-
Seabed mapping is a common application for marine robots, and it is often framed as a coverage path planning problem in robotics. During a robot-based survey, the coverage of perceptual sensors (e.g., cameras, LIDARS and sonars) changes, especially in underwater environments. Therefore, online path planning is needed to accommodate the sensing changes in order to achieve the desired coverage ratio. In this paper, we present a sensing confidence model and a uncertainty-driven sampling-based online coverage path planner (SO-CPP) to assist in-situ robot planning for seabed mapping and other survey-type applications. Different from conventional lawnmower pattern, the SO-CPP will pick random points based on a probability map that is updated based on in-situ sonar measurements using a sensing confidence model. The SO-CPP then constructs a graph by connecting adjacent nodes with edge costs determined using a multi-variable cost function. Finally, the SO-CPP will select the best route and generate the desired waypoint list using a multi-variable objective function. The SO-CPP has been evaluated in a simulation environment with an actual bathymetric map, a 6-DOF AUV dynamic model and a ray-tracing sonar model. We have performed Monte Carlo simulations with a variety of environmental settings to validate that the SO-CPP is applicable to a convex workspace, a non-convex workspace, and unknown occupied workspace. So-CPP is found outperform regular lawnmower pattern survey by reducing the resulting traveling distance by upto 20%. Besides that, we observed that the prior knowledge about the obstacles in the environment has minor effects on the overall traveling distance. In the paper, limitation and real-world implementation are also discussed along with our plan in the future.more » « less
-
null (Ed.)This paper presents a novel framework for training convolutional neural networks (CNNs) to quantify the impact of gradual and abrupt uncertainties in the form of adversarial attacks. Uncertainty quantification is achieved by combining the CNN with a Gaussian process (GP) classifier algorithm. The variance of the GP quantifies the impact on the uncertainties and especially their effect on the object classification tasks. Learning from uncertainty provides the proposed CNN-GP framework with flexibility, reliability and robustness to adversarial attacks. The proposed approach includes training the network under noisy conditions. This is accomplished by comparing predictions with classification labels via the Kullback-Leibler divergence, Wasserstein distance and maximum correntropy. The network performance is tested on the classical MNIST, Fashion-MNIST, CIFAR10 and CIFAR 100 datasets. Further tests on robustness to both black-box and white-box attacks are also carried out for MNIST. The results show that the testing accuracy improves for networks that backpropogate uncertainty as compared to methods that do not quantify the impact of uncertainties. A comparison with a state-of-art Monte Carlo dropout method is also presented and the outperformance of the CNN-GP framework with respect to reliability and computational efficiency is demonstrated.more » « less
-
Creating soft robots with sophisticated, autonomous capabilities requires these systems to possess reliable, on-line proprioception of 3D configuration through integrated soft sensors. We present a framework for predicting a soft robot’s 3D configuration via deep learning using feedback from a soft, proprioceptive sensor skin. Our framework introduces a kirigami-enabled strategy for rapidly sensorizing soft robots using off-the-shelf materials, a general kinematic description for soft robot geometry, and an investigation of neural network designs for predicting soft robot configuration. Even with hysteretic, non-monotonic feedback from the piezoresistive sensors, recurrent neural networks show potential for predicting our new kinematic parameters and, thus, the robot’s configuration. One trained neural network closely predicts steady-state configuration during operation, though complete dynamic behavior is not fully captured. We validate our methods on a trunk-like arm with 12 discrete actuators and 12 proprioceptive sensors. As an essential advance in soft robotic perception, we anticipate our framework will open new avenues towards closed loop control in soft robotics.more » « less
