skip to main content


Search for: All records

Award ID contains: 2034123

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. We examine a novel setting in which two parties have partial knowledge of the elements that make up a Markov Decision Process (MDP) and must cooperate to compute and execute an optimal policy for the problem constructed from those elements. This situation arises when one party wants to give a robot some task, but does not wish to divulge those details to a second party-while the second party possesses sensitive data about the robot's dynamics (information needed for planning). Both parties want the robot to perform the task successfully, but neither is willing to disclose any more information than is absolutely necessary. We utilize techniques from secure multi-party computation, combining primitives and algorithms to construct protocols that can compute an optimal policy while ensuring that the policy remains opaque by being split across both parties. To execute a split policy, we also give a protocol that enables the robot to determine what actions to trigger, while the second party guards against attempts to probe for information inconsistent with the policy's prescribed execution. In order to improve scalability, we find that basis functions and constraint sampling methods are useful in forming effective approximate MDPs. We report simulation results examining performance and precision, and assess the scaling properties of our Python implementation. We also describe a hardware proof-of-feasibility implementation using inexpensive physical robots, which, being a small-scale instance, can be solved directly. 
    more » « less
    Free, publicly-accessible full text available December 13, 2024
  2. We present an incremental scalable motion planning algorithm for finding maximally informative trajectories for decentralized mobile robots. These robots are deployed to observe an unknown spatial field, where the informativeness of observations is specified as a density function. Existing works that are typically restricted to discrete domains and synchronous planning often scale poorly depending on the size of the problem. Our goal is to design a distributed control law in continuous domains and an asynchronous communication strategy to guide a team of cooperative robots to visit the most informative locations within a limited mission duration. Our proposed Asynchronous Information Gathering with Bayesian Optimization (AsyncIGBO) algorithm extends ideas from asynchronous Bayesian Optimization (BO) to efficiently sample from a density function. It then combines them with decentralized reactive motion planning techniques to achieve efficient multi-robot information gathering activities. We provide a theoretical justification for our algorithm by deriving an asymptotic no-regret analysis with respect to a known spatial field. Our proposed algorithm is extensively validated through simulation and real-world experiment results with multiple robots. 
    more » « less
    Free, publicly-accessible full text available May 1, 2024
  3. Digital technology has evolved towards a new way of processing information: web searches, social platforms, internet forums, and video games have substituted reading books and writing essays. Trainers and educators currently face the challenge of providing natural training and learning environments for digital-natives. In addition to physical spaces, effective training and education require virtual spaces. Digital twins enable trainees to interact with real hardware in virtual training environments. Interactive real-world elements are essential in the training of robot operators. A natural environment for the trainee supports an interesting learning experience while including enough professional substances. We present examples of how virtual environments utilizing digital twins and extended reality can be applied to enable natural and effective robotic training scenarios. Scenarios are validated using cross-platform client devices for extended reality implementations and safety training applications.

     
    more » « less
  4. Prediction and estimation of phenomena of interest in aquatic environments are challenging since they present complex spatio-temporal dynamics. Over the past few decades, advances in machine learning and data processing contributed to ocean exploration and sampling using autonomous robots. In this work, we formulate a reinforcement learning framework to estimate spatio-temporal fields modeled by partial differential equations. The proposed framework addresses problems of the classic methods regarding the sampling process to determine the path to be used by the agent to collect samples. Simulation results demonstrate the applicability of our approach and show that the error at the end of the learning process is close to the expected error given by the fitting process due to added noise. 
    more » « less
  5. The use of underwater robot systems, including Autonomous Underwater Vehicles (AUVs), has been studied as an effective way of monitoring and exploring dynamic aquatic environments. Furthermore, advances in artificial intelligence techniques and computer processing led to a significant effort towards fully autonomous navigation and energy-efficient approaches. In this work, we formulate a reinforcement learning framework for long-term navigation of underwater vehicles in dynamic environments using the techniques of tile coding and eligibility traces. Simulation results used actual oceanic data from the Regional Ocean Modeling System (ROMS) data set collected in Southern California Bight (SCB) region, California, USA 
    more » « less
  6. Many studies suggest that water quality parameters can be estimated by applying statistical and machine learning methods using remote sensing or in-situ data. However, identifying best practices for implementing solutions appears to be done on a case-by-case basis. In our case, we have in-situ data that covers a large period, but only small areas of Biscayne Bay, Florida. In this paper, we combine available in-situ data with remote sensing data captured by Landsat 8 OLI-TIRS Collection 2 Level 2(L8), Sentinel-2 L2A(S2), and Sentinel-3 OLCI L1B(S3). The combined data set is for use in a water quality parameter estimation application. Our contributions are two-fold. First, we present a pipeline for data collection, processing, and co-location that results in a usable data set of combined remote sensing and in-situ data. Second, we propose a classification model using the combined data set to identify areas of interest for future data collection missions based on chlorophyll-a in-situ measurements. To further prove our methodology, we conduct a data collection mission using one of the predicted paths from our model. 
    more » « less
  7. Given sensor units distributed throughout an environment, we consider the problem of consolidating readings into a single coherent view when sensors wish to limit knowledge of their specific readings. Standard fusion methods make no guarantees about what curious participants may learn. For applications where privacy guarantees are required, we introduce a fusion approach that limits what can be inferred. First, it forms an aggregate stream, oblivious to the underlying sensor data, and then evaluates that stream on a combinatorial filter. This is achieved via secure multi-party computation techniques built on cryptographic primitives, which we extend and apply to the problem of fusing discrete sensor signals. We prove that the extensions preserve security under the model of semi-honest adversaries. Also, for a simple target tracking case study, we examine a proof-of-concept implementation: analyzing the (empirical) running times for components in the architecture and suggesting directions for future improvement. 
    more » « less
  8. null (Ed.)
  9. null (Ed.)
  10. null (Ed.)
    There is a need for semi-autonomous systems capable of performing both automated tasks and supervised maneuvers. When dealing with multiple robots or robots with high complexity (such as humanoids), we face the issue of effectively coordinating operators across robots. We build on our previous work to present a methodology for designing trajectories and policies for robots such that a few operators can supervise multiple robots. Specifically, we: (1) Analyze the complexity of the problem, (2) Design a procedure for generating policies allowing operators to oversee many robots, (3) Present a method for designing policies and robot trajectories to allow operators to oversee multiple robots, and (4) Include both simulation and hardware experiments demonstrating our methodologies. 
    more » « less