Conference Presentations
AAAI Spring Symposia 2022
The ability of AI agents and architectures to detect and adapt to sudden changes in their environments remains an outstanding challenge. In the context of multi-agent games, the agent may face novel situations where the rules of the game, the available actions, the environment dynamics, the behavior of other agents, as well as the agent’s goals suddenly change. In this paper, we introduce an architecture that allows agents to detect novelties, characterize those novelties, and build an appropriate adaptive model to accommodate them. Our agent utilizes logic and reasoning (specifically, Answer Set Programming) to characterize novelties into different categories, as to enable the agent to adapt to the novelty while maintaining high performance in the game. We demonstrate the effectiveness of the proposed agent architecture in a multi-agent imperfect information board game, Monopoly. We measure the success of the architecture by comparing our method to heuristics, and vanilla Monte-Carlo Tree Search approaches. Our results indicate precise novelty detection, and significant improvements in the performance of agents utilizing the novelty handling architecture.
Joint Mathematics Meeting - January 2019
In the past few years, several major cybersecurity attacks on supervisory control and data acquisition (SCADA) devices have been reported. Such attacks can result in damages to the economy and have an impact on society. In 2010, Ten et al. presented an attack tree mode of impact analysis. We have implemented the attack tree structure developed by Ten, in concert with typical financial loss data to implement Monte Carlo techniques to generate a new cost-benefit analysis of various security improvement scenarios. Time to attack is modeled as an exponentially distributed random variable obtained via maximum likelihood analysis; financial losses are modeled using regression to generalized logistic functions via gradient descent. Under these conditions, hypothetical future losses are simulated for a variety of intrusion scenarios and improvement schemes. Our model incorporates budgetary constraints in an effort to advise the prioritization of system improvements, and we compare the results of genetic and differential evolution algorithms in determining an optimal budget allocation.
American Association for the Advancement of Science - February 2019
Background: The negative photoresist SU-8 has been adopted in a wide range of applications, such as the fabrication of MEMS and microfluidics. Despite extensive studies on the optical properties of SU-8, the exposure dependence on the wavelength of monochromatic ultraviolet (UV) light sources has not been reported. The work presented here investigates such dependency by exposing SU-8 using three single-wavelength UV-light sources of differing wavelengths.
Methods: Experiments were conducted using the low-cost (under $4000) in-house photolithography system with an adjustable light source. Three light sources, implemented with single-wavelength UV LEDs at the wavelengths of 365-, 385-, and 405-nm, were used to study the exposure dependence on the wavelength. The resolution test for each wavelength was executed with the feedback dosage control incorporated in the system.
Results: Exposure under 365- and 385-nm UV-light produced vertical sidewalls with a positive taper angle of fewer than 3.64 degrees. A resolution of 19um was achieved using 365nm UV-light in a 30- and 50-um thick photoresist with a dosage of 250 and 330 mJ/cm2, respectively. For the 385nm UV-light, 30um resolution was obtained with a dosage of 4750 mJ/cm2 for 30um thickness and 7100mJ/cm2 for 50um thickness. Exposure under 405nm UV-light yielded a large positive taper angle of 7.54 degrees in 50um thick photoresist, which was due to significantly lower absorbance at this wavelength.
Conclusions: The exposure of SU-8, in terms of resolution, optimal dosage, and sidewall profile, was found to be extremely sensitive to the wavelength of the UV-light between 365- and 405-nm. When the wavelength increased from 365- to 385-nm , the resolution was worsened by 57.89% (from 19um to 30um), the exposure dosage increased by a factor of 19, but the sidewall remained vertical. Changing the wavelength from 385- to 405-nm resulted in a positive taper angle of 7.54 degrees. Such non-vertical sidewall profile only observed at 405nm was attributed to the lowest absorbance as well as the greatest difference in absorption coefficient between unexposed and exposed photoresist at this wavelength.