Somchaya Liemhetcharat
Research

My long-term research goal is to have multiple robots coordinating and performing some useful task with minimal human intervention. The multi-robot team would need to gather resources (e.g., power from sunlight, materials for repairs), and ensure that the task remains continuously executed even when some robots need down-time (e.g., to rest, recharge and repair).

To accomplish my long-term goal, there are many research areas that I am interested in. Click here for a list of projects that I did during my Ph.D. at Carnegie Mellon University. Below is a list of the research areas I am currently pursuing and/or looking for students to work on. Feel free to contact me via email: som [at] somchaya -dot- org, if you are interested in working on anything below.

Ad Hoc Teams
As research in robotics and artificial intelligence progresses, robots will become more intelligent and autonomous. However, since different universities and research institutes create different robots, these robots may or may not be able to coordinate well when put together in a team. An ad hoc team is one where the team members may not have coordinated before, so their team performance may be unknown. This research area is an extension of my Ph.D thesis on Representation, Planning, and Learning of Dynamic Ad Hoc Robot Teams. The focus of this research is on modeling team performance and forming a multi-robot team, i.e., modeling the interactions of the robots, and selecting which robots should be part of the team, in order to maximize team performance.

Multi-Robot Coordination
Besides modeling how ad hoc teams will perform, I am also interested in research on improving the coordination among robots in a team. Multi-robot coordination is a broad field, and my areas of interest include: continuous task execution, coordination strategies, task allocation, and multi-robot planning. I am particularly interested in heterogeneous robot teams, where robots have different hardware and capabilities. I am interested in algorithms that can run real-time on a robot. My view is that robots should be autonomous and perform most (if not all) computation on-board without requiring external sensors or processors. External sensors and processors should be used when available to improve a robot's performance, but the robot should have a baseline that allows it to act autonomously. This area is an extension of my work in LANdroids and Robot Soccer.

The video below shows the undergraduate project on Vision-based Autonomous Control and Navigation of an Unmanned Aerial Vehicle (UAV), and the next step is to enable multiple UAVs and ground robots to coordinate.

Human-Robot Teams
In general, robots are well-suited for dirty, dangerous, and dull tasks. Robots can perform tasks with high precision and consistency, while humans are good at big picture planning. This research area focuses on the synergy between humans and robots, to maximize the overall task performance. Some topics of interest are: modeling the boundaries of the robot's and human's abilities, information-sharing between robots and humans, effective command and coordination strategies for a human-robot team, and synergizing the human's and robot's abilities to perform a task.

Game-Playing Artificial Intelligence
Learning to play a game has many similarities to learning to perform a task. I am interested in creating AI agents and robots that can play games: real-life games such as soccer, and computer games such as Civilization, Starcraft, and World of Warcraft. The focus of this research area is on learning from experience and simulations to improve overall performance, and to leverage on research in the other areas above. This area is an extension of my work in Robot Soccer.

Nurturing the Next Generation
I was fortunate to have many opportunities during my studies that nurtured my interest in Science and Robotics, and I want to create such opportunities for the next generation. As such, I am generally happy to work with self-motivated students who want to learn more about artificial intelligence and robotics. Also, I believe that robotics is a great avenue for generating interest in Science, Technology, Engineering, and Mathematics (STEM). Hence, I am very interested in developing curriculum to teach STEM topics using robots, and to teach robotics to encourage interest in STEM. This area is an extension of my work in the FIRE project at CMU, and the internships during my Ph.D.