KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/kmitl/1

Browse

Search Results

Now showing 1 - 5 of 5

Some of the metrics are blocked by your
consent settings
n-LIPO: Framework for Diverse Cooperative Agent Generation Using Policy Compatibility
(2025-01-01)
Charakorn, Rujikorn
;
Manoonpong, Poramate
;
Dilokthanakul, Nat
Diverse training partners in multiagent tasks are crucial for training a robust and adaptable cooperative agent. Prior methods often rely on state-action information to diversify partners’ behaviors, but this can lead to minor variations instead of diverse behaviors and solutions. We address this limitation by introducing a novel training objective based on “policy compatibility.” Our method learns diverse behaviors by encouraging agents within a team to be compatible with each other while being incompatible with agents from other teams. We theoretically prove that incompatible policies are inherently dissimilar, allowing us to use policy compatibility as a proxy for diversity. We call this method learning incompatible policies for n -player cooperative games (n-LIPO). We propose to further diversify individual policies by incorporating a mutual information objective using state-action information. We empirically demonstrate that n-LIPO effectively generates diverse joint policies in various two-player and multi-player cooperative environments. In a complex cooperative task, two-player multi-recipe Overcooked, we find that n-LIPO generates a population of behaviorally diverse partners. These populations are then used to train robust generalist agents that can generalize better than using baseline populations. Finally, we demonstrate that n-LIPO can be applied to a high-dimensional StarCraft multiagent challenge (SMAC) multiplayer cooperative environment to discover diverse winning strategies when only a single goal exists.
4
Some of the metrics are blocked by your
consent settings
ThaiScamBench: Toward a Benchmark Dataset for Scam and Phishing Detection in the Thai Language
(2025-01-01)
Chalaemwongwan, Nutthakorn
Recent reports from Thai financial regulators reveal a sharp increase in online scams, resulting in major financial damage each year. While English-language research has made considerable progress, Thai-language scam detection remains underexplored and lacks standardized benchmarks. The study establishes *ThaiScamBench (pilot)*, a curated corpus of 1,750 Thai messages labeled as scam or legitimate across seven categories. Reproducible baselines (Logistic Regression, Linear SVM) and a rigorous evaluation protocol are provided to address class imbalance, Thai-English code-switching, and adversarial obfuscation. All scam texts contained URLs; none appeared in legitimate messages, revealing dataset bias, with all scam samples containing a URL and none among legitimate texts, artificially inflating headline metrics. We addressed this issue by introducing URL-masked and domain hold-out evaluations. We outline our roadmap toward version 1.0 (around 50k messages), focusing on dataset scaling and robust benchmarking for Thai scam detection, adversarial stress tests, and PDPA-compliant release artifacts. *ThaiScamBench∗ establishes the first standardized evaluation for Thai scam detection, enabling transparent comparison and artifact-conscious evaluation. This benchmark provides a reproducible foundation for robust Thai-language scam detection.
Some of the metrics are blocked by your
consent settings
Parallel Active Power Filter Based on Source Current Detection for Antiparallel Resonance with Robustness to Parameter Variations in Power Systems
(2019-02-01)
Yeetum, Wanno
;
Kinnares, Vijit
In previous research, the variable conductance of a parallel active power filter has been used for antiparallel resonance with parameter variations in power systems. Its disadvantages are many sensors, complicated harmonic detection, and complexity of the algorithm for successful accomplishment in antiparallel resonance. In order to eliminate such drawbacks, this paper presents the parallel active power filter based on source current detection for antiparallel resonance with robustness to parameter variations for power factor correction using capacitor banks. The proposed method employs only single harmonic detector and an open-loop shaping design for robustness to variations of source impedance and a sudden change of capacitor bank values for power factor correction. The proposed antiparallel resonance controller can be readily applied to both large-scale and laboratory-scale power systems without any change. The large-scale simulation and the laboratory-scale experiment results confirm the validity, the accuracy, and good performance for antiparallel resonance, harmonic suppression, and robustness to parameter variations in power systems.
Some of the metrics are blocked by your
consent settings
Optimal and robust strategies for freeway traffic management under demand and supply uncertainties: An overview and general theory
(2014-01-01)
Zhong, R. X.
;
Sumalee, A.
;
Pan, T. L.
;
Lam, William H.K.
This paper investigates optimal decision-making for traffic management under demand and supply uncertainties by stochastic dynamic programming. Traffic flow dynamics under demand and supply uncertainties is described by a simplified version of the stochastic cell transmission model. Based on this model, the optimal traffic management problem is analysed wherein the existence of solution is guaranteed by verifying the well-posed condition. An analytical optimal control law is derived in terms of a set of coupled generalised recursive Riccati equations. As optimal control laws may be fragile with respect to model misspecification, a robust (optimal) decision-making law that aims to act robust with respect to the parameter misspecification in the traffic flow model (which can be originated from model calibration), and to attenuate the effect of disturbances in freeway networks (wherein demand uncertainty is usually regarded as a kind of disturbance) is proposed. Conventionally, network uncertainties have been considered to induce negative effects on traffic management in transportation literature. In contrast, the proposed methodology outlines an interesting issue that is to make benefit (or trade-off) from the inherent network uncertainties. Finally, some practical issues in traffic management that can be addressed by extending the current framework are briefly discussed. © 2014 Hong Kong Society for Transportation Studies Limited.
Some of the metrics are blocked by your
consent settings
Robustness of 3D indoor localization based on fingerprint technique in wireless sensor networks
(2013-09-02)
Chuenurajit, Thanapong
;
Phimmasean, Sisongkham
;
Cherntanomwong, Panarat
The simple approaches of an indoor localization have been continuously developed and extensively published. In order to achieve a challenge of the effective indoor localization, the expected localization system should be a high efficiency such as high accuracy, simple, robustness and effective system. For the indoor localization, the researches have been supported by the advancement of the sensor node technology. One application of this technology is wireless sensor networks-based indoor localization. The simple approaches of 2-dimensional (2D) indoor localization have been widely proposed. Since a realistic system includes complicated terrain and different environment, 3-dimensional (3D) consideration is more suitable to be applied in the real life. This paper proposes an approach of 3D indoor localization based on ZigBee standard. Fingerprint technique-based received signal strength indicator (RSSI) is employed. Due to fluctuating signals in indoor environment, robustness of fingerprint technique will be proposed in order to solve the propagation mechanisms. The k-Nearest Neighbor using Euclidean distance is utilized as the pattern matching algorithm. For the case study, 8 reference nodes and 1 target node are stationary placed on a bookshelf in clean and human body's effect environments. The expected errors of estimated target locations should not be more than 36 cm (each level height). From the results, we can acquire an acceptable accuracy. It shows that our system can be applied in the real application. © 2013 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results