An official website of the United States government
We introduce temporal multimodal multivariate learning, a new family of decision making models that can indirectly learn and transfer online information from simultaneous observations of a probability distribution with more than one peak or more than one outcome variable from one time stage to another. We approximate the posterior by sequentially removing additional uncertainties across different variables and time, based on data-physics driven correlation, to address a broader class of challenging time-dependent decision-making problems under uncertainty. Extensive experiments on real-world datasets ( i.e., urban traffic data and hurricane ensemble forecasting data) demonstrate the superior performance of the proposed targeted decision-making over the state-of-the-art baseline prediction methods across various settings.
more »
« less
A Computational Model of Hurricane Evacuation Decision.
Hurricanes are devastating natural disasters. In deciding how to respond to a hurricane, in particular whether and when to evacuate, a decision-maker must weigh often highly uncertain and contradic- tory information about the future path and intensity of the storm. To effectively plan to help people during a hurricane, it is crucial to be able to predict and understand this evacuation decision. To this end, we propose a computational model of human sequential decision-making in response to a hurricane based on a Partial Ob- servable Markov Decision Process (POMDP) that models concerns, uncertain beliefs about the hurricane, and future information. We evaluate the model in two ways. First, hurricane data from 2018 was used to evaluate the model’s predictive ability on real data. Second, a simulation study was conducted to qualitatively evaluate the sequential aspect of the model to illustrate the role that the acquisition of future, more accurate information can play on cur- rent decision-making. The evaluation with 2018 hurricane season data shows that our proposed features are significant predictors and the model can predict the data well, within and across distinct hurricane datasets. The simulation results show that, across dif- ferent setups, our model generates predictions on the sequential decisions making aspect that align with expectations qualitatively and suggests the importance of modeling information.
more »
« less
- Award ID(s):
- 1638234
- PAR ID:
- 10176161
- Date Published:
- Journal Name:
- 19th International Conference on Autonomous Agents and MultiAgent Systems.
- Page Range / eLocation ID:
- 2062--2064
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Storm surge and evacuation traffic under the observed track of Hurricane Michael (2018) showed clear accessibility and evacuation challenges for Panama City, Florida although the city was not hit directly. Since a possible Hurricane Michael track within National Hurricane Center (NHC)'s forecasted hurricane cone was Panama City, this paper tries to answer the following questions: What if Hurricane Michael hit Panama City directly? How would the special needs populations and their accessibility to Special Needs Shelters (SpNS) be impacted, and what could have been done to alleviate this impact? A previously validated storm surge model was used to predict storm surge inundations under this different hurricane track. Based on the impact of these coastal inundations, a GIS-based optimization methodology was developed to evaluate the accessibility and siting of special needs shelters. Results indicate that if Hurricane Michael had shifted to Panama City in 2018, most of the coastal region of Panama City would have been inundated, compelling residents to evacuate. The possible landfall of Michael in this simulation would also lead to a maximum storm surge of 5 to 6 m on the coast, which is above FEMA's 100-year flood elevation. In addition, the only evacuation route out of Panama City area, when the bridges with their access roads were flooded, was US 231. This would have been life-threatening since there is only one SpNS in the north of the city accessible by this roadway. The proposed analysis studies the accessibility of this SpNS shelter and provides a reasonable approach for SpNS shelter siting or repurposing regular shelters for this purpose based on the hypothesized travel time most likely to be experienced on roadway networks based on the impact of Hurricane Michael. Emergency plans can be updated by the results of this optimization model, which can locate additional sites or shelter locations while minimizing the travel costs and integrating the impact of storm surge modeling and transportation accessibility analysis.more » « less
-
Predictive monitoring—making predictions about future states and monitoring if the predicted states satisfy requirements—offers a promising paradigm in supporting the decision making of Cyber-Physical Systems (CPS). Existing works of predictive monitoring mostly focus on monitoring individual predictions rather than sequential predictions. We develop a novel approach for monitoring sequential predictions generated from Bayesian Recurrent Neural Networks (RNNs) that can capture the inherent uncertainty in CPS, drawing on insights from our study of real-world CPS datasets. We propose a new logic named Signal Temporal Logic with Uncertainty (STL-U) to monitor a flowpipe containing an infinite set of uncertain sequences predicted by Bayesian RNNs. We define STL-U strong and weak satisfaction semantics based on whether all or some sequences contained in a flowpipe satisfy the requirement. We also develop methods to compute the range of confidence levels under which a flowpipe is guaranteed to strongly (weakly) satisfy an STL-U formula. Furthermore, we develop novel criteria that leverage STL-U monitoring results to calibrate the uncertainty estimation in Bayesian RNNs. Finally, we evaluate the proposed approach via experiments with real-world CPS datasets and a simulated smart city case study, which show very encouraging results of STL-U based predictive monitoring approach outperforming baselines.more » « less
-
Learners that are exposed to the same training data might generalize differently due to differing inductive biases. In neural network models, inductive biases could in theory arise from any aspect of the model architecture. We investigate which architectural factors affect the generalization behavior of neural sequence-to-sequence models trained on two syntactic tasks, English question formation and English tense reinflection. For both tasks, the training set is consistent with a generalization based on hierarchical structure and a generalization based on linear order. All architectural factors that we investigated qualitatively affected how models generalized, including factors with no clear connection to hierarchical structure. For example, LSTMs and GRUs displayed qualitatively different inductive biases. However, the only factor that consistently contributed a hierarchical bias across tasks was the use of a tree-structured model rather than a model with sequential recurrence, suggesting that human-like syntactic generalization requires architectural syntactic structure.more » « less
-
Human navigation simulation is critical to many civil engineering tasks and is of central interest to the simulation community. Most human navigation simulation approaches focus on the classic psychology evidence, or assumptions that still need further proofs. The overly simplified and generalized assumption of navigation behaviors does not highlight the need of capturing individual differences in spatial cognition and navigation decision-making, or the impacts of diverse ways of spatial information display. This study proposes the visual attention patterns in floorplan review to be a stronger predictor of human navigation behaviors. To set the theoretical foundation, a Virtual Reality (VR) experiment was performed to test if visual attention patterns during spatial information review can predict the quality of spatial memory, and how the relationship is affected by the diverse ways of information display, including 2D, 3D and VR. The results set a basis for future prediction model developments.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
