An official website of the United States government
We present a hierarchical control approach for maneuvering an autonomous vehicle (AV) in tightly-constrained environments where other moving AVs and/or human driven vehicles are present. A two-level hierarchy is proposed: a high-level data-driven strategy predictor and a lower-level model-based feedback controller. The strategy predictor maps an encoding of a dynamic environment to a set of high-level strategies via a neural network. Depending on the selected strategy, a set of time-varying hyperplanes in the AV’s position space is generated online and the corresponding halfspace constraints are included in a lower-level model-based receding horizon controller. These strategy-dependent constraints drive the vehicle towards areas where it is likely to remain feasible. Moreover, the predicted strategy also informs switching between a discrete set of policies, which allows for more conservative behavior when prediction confidence is low. We demonstrate the effectiveness of the proposed data-driven hierarchical control framework in a two-car collision avoidance scenario through simulations and experiments on a 1/10 scale autonomous car platform where the strategy-guided approach outperforms a model predictive control baseline in both cases.
more »
« less
This content will become publicly available on May 6, 2026
Federated Privacy-Preserving Strategy for Generalizing Soft-Failure Localization in Multi-Carrier Optical Networks
We propose a privacy-preserving strategy based on federated learning to localize soft failures in multi-carrier optical networks using a self-supervised approach on unlabeled data. Evaluations conducted on data from a testbed demonstrate the effectiveness of the proposed strategy.
more »
« less
- Award ID(s):
- 2210384
- PAR ID:
- 10646460
- Publisher / Repository:
- IEEE
- Date Published:
- ISSN:
- 2995-0686
- ISBN:
- 978-3-903176-67-6
- Page Range / eLocation ID:
- 1 to 3
- Format(s):
- Medium: X
- Location:
- Pisa, Italy
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
This paper investigates the idea of introducing learning algorithms into parking guidance and information systems that employ a central server, in order to provide estimated optimal parking searching strategies to travelers. The parking searching process on a network with uncertain parking availability can naturally be modeled as a Markov Decision Process (MDP). Such an MDP with full information can easily be solved by dynamic programming approaches. However, the probabilities of finding parking are difficult to define and calculate, even with accurate occupancy data. Learning algorithms are suitable for addressing this issue. The central server collects data from numerous travelers’ parking search experiences in the same area within a time window, computes approximated optimal parking searching strategy using a learning algorithm, and distributes the strategy to travelers. We propose an algorithm based on Q-learning, where the topology of the underlying transportation network is incorporated. This modification allows us to reduce the size of the problem dramatically, and thus the amount of data required to learn the optimal strategy. Numerical experiments conducted on a toy network show that the proposed learning algorithm outperforms the nearest-node greedy search strategy and the original Q-learning algorithm. Sensitivity analysis regarding the desired amount of training data is also performed.more » « less
-
Current Hi-C analysis approaches focus on uniquely mapped reads and little research has been carried out to include multi-mapping reads, which leads to a lack of biological signals from DNA repetitive regions. We propose a heuristic strategy to assign multi-mapping reads to loci according to the distance to their closest restriction enzyme cutting sites. We demonstrate that the heuristic strategy can rescue multi-mapping reads thus enhance the quality of Hi-C data. Compared with mHi-C, it not only improves replicate reproducibility in the same cell type, but also maintains the difference between replicates of different cell types. Moreover, the strategy identifies much more common statistically significant chromatin interactions between Hi-C experiments of different restriction enzymes and has a huge advantage on computing resources. Therefore, the heuristic strategy can be used to enhance Hi-C data by utilizing multi-mapping reads.more » « less
-
We propose and evaluate a lightweight strategy for tracing code that can be efficiently taught to novice programmers, building off of recent findings on "sketching" when tracing. This strategy helps novices apply the syntactic and semantic knowledge they are learning by encouraging line-by-line tracing and providing an external representation of memory for them to update. To evaluate the effect of teaching this strategy, we conducted a block-randomized experiment with 24 novices enrolled in a university-level CS1 course. We spent only 5-10 minutes introducing the strategy to the experimental condition. We then asked both conditions to think-aloud as they predicted the output of short programs. Students using this strategy scored on average 15% higher than students in the control group for the tracing problems used the study (p<0.05). Qualitative analysis of think-aloud and interview data showed that tracing systematically (line-by-line and "sketching" intermediate values) led to better performance and that the strategy scaffolded and encouraged systematic tracing. Students who learned the strategy also scored on average 7% higher on the course midterm. These findings suggest that in <1 hour and without computer-based tools, we can improve CS1 students' tracing abilities by explicitly teaching a strategy.more » « less
-
Local outlier techniques are known to be effective for detecting outliers in skewed data, where subsets of the data exhibit diverse distribution properties. However, existing methods are not well equipped to support modern high-velocity data streams due to the high complexity of the detection algorithms and their volatility to data updates. To tackle these shortcomings, we propose local outlier semantics that operate at an abstraction level by leveraging kernel density estimation (KDE) to effectively detect local outliers from streaming data. A strategy to continuously detect top-N KDE-based local outliers over streams is designed, called KELOS – the first linear time complexity streaming local outlier detection approach. The first innovation of KELOS is the abstract kernel center-based KDE (aKDE) strategy. aKDE accurately yet efficiently estimates the data density at each point – essential for local outlier detection. This is based on the observation that a cluster of points close to each other tend to have a similar influence on a target point’s density estimation when used as kernel centers. These points thus can be represented by one abstract kernel center. Next, the KELOS’s inlier pruning strategy early prunes points that have no chance to become top-N outliers. This empowers KELOS to skip the computation of their data density and of the outlier status for every data point. Together aKDE and the inlier pruning strategy eliminate the performance bottleneck of streaming local outlier detection. The experimental evaluation demonstrates that KELOS is up to 6 orders of magnitude faster than existing solutions, while being highly effective in detecting local outliers from streaming data.more » « less
