Research

We have developed a sleep interpretation AI that not only translates complex sleep data into understandable reports using language models but also powers a chatbot service that provides users with summaries, insights, and personalized improvement actions for better sleep. This system continuously evolves, incorporating user feedback to refine the model for more accurate and helpful advice.

[Sleep Routine App]

To enhance the realism in autonomous driving simulations, we address the shortcomings of current Learning from Demonstration (LfD) methods, which often result in unrealistic driving behaviors like collisions or veering off the road. We introduce Symphony, a novel solution that combines standard driving policies with a parallel beam search to refine these policies in real-time. This beam search eliminates less realistic outcomes based on a discriminator's assessment but risks reducing the diversity of driving behaviors by encouraging mode collapse. Symphony counters this by adopting a hierarchical structure that separates the processes of generating goals and conditioning behaviors on these goals. This ensures a broad spectrum of realistic behaviors is maintained, preventing loss of diversity during adversarial training or through beam search pruning. Our validation with both proprietary and open Waymo datasets demonstrates that Symphony achieves significantly more realistic and diverse agent behaviors compared to existing baselines.

[ICRA 2022]

One way to accelerate the coordination effect is to enable multiple agents to communicate with each other in a distributed manner and behave as a group. In this research, we study a practical scenario when (i) the communication bandwidth is limited and (ii) the agents share the communication medium so that only a restricted number of agents are able to simultaneously use the medium, as in the state-of-the-art wireless networking standards. This calls for a certain form of communication scheduling. In that regard, we propose a multi-agent deep reinforcement learning framework, called SchedNet, in which agents learn how to schedule themselves, how to encode the messages, and how to select actions based on received messages. 

[ICLR 2019]

How can we realize agents' cooperation over distributed execution environments without any communication via a value-based multi-agent reinforcement learning? We proposed the way of factorizing the global Q-function into individual ones to achieve the goal. 

[ICML 2019]

The other way to accelerate the cooperation among the agents is shaping reward. The agents are trained based on the reward they get after taking actions. How to valuate the agent's policy determines how the agents will be trained. We also make the trainer learn how to re-shape the reward such that multiple agents are cooperate each other. 

[AAMAS 2020]

We introduce Fog Operating System (FogOS), a fog computing architecture for IoT services. We take the perspective of designing an operating system, practicing the architectural lessons from the long history of operating systems. We focus on addressing the challenges raised by (i) diversity and heterogeneity of IoT services and (ii) edge devices that are owned by individuals and different owners, and presenting how FogOS is designed to effectively and efficiently provide and manage such IoT services. We provide two use cases to demonstrate FogOS in action.

[Comm. Magazine 2017]

One of the crucial factors towards the success of fog computing is how to incentivize the individual users’ edge resources, thereby opening the era of user-participated fog computing. In this research, we provide an economic analysis of such user-oriented fog computing by modeling a market consisting of ISP (Infrastructure and Service Provider), SUs (end Service Users), and EROs (Edge Resource Owners) as a non-cooperative game. In this market, ISP, which provides a platform of fog computing, behaves as a mediator or a broker to lease the edge resources from EROs and provide various services to SUs. 

[ICC 2018], [IEEE Transaction on Mobile Computing 2019]

We propose a new sensor MAC protocol, called Bird-MAC, which is highly energy efficient in the applications where sensors periodically report monitoring status with a very low rate, as in structural health monitoring and static environmental monitoring. Two key design ideas of Bird-MAC are: (a) no need of early-wake-up of transmitters and (b) taking the right balance between synchronization and coordination costs. The idea (a) is possible by allowing a node (whether it is a transmitter or receiver) to wake up just with its given wake-up schedule, and letting a late bird (which wakes up later) notify its wake-up status to its corresponding early bird (which wakes up earlier), where the early bird just infrequently waits (i.e., nods) for the late bird’s wake-up signal. The idea (b) is realized by designing Bird-MAC to be placed in a scheme between purely synchronous and asynchronous schemes.

[IEEE SECON 2017], [IEEE Transaction on Mobile Computing 2019]