Projects are listed below in reverse chronological order.


Short-term Conflict Resolution for Unmanned Aircraft Management (paper, slides, code)

H. Y. Ong and M. J. Kochenderfer

  • We propose a set of controllers to solve the stochastic short-term conflict avoidance problem, in which a traffic management system supervises a collection of aircraft. The controllers generate advisories for each aircraft, and are based on decomposing a Markov decision process and fusing solutions to its subproblems online. (Published in 34th IEEE/AIAA Digital Avionics Systems Conference, Prague, Czech Republic, September 2015; Best Student Paper.)

Cooperative Collision Avoidance via Proximal Message Passing (paper, slides)

H. Y. Ong and J. C. Gerdes

  • We propose a distributed model predictive controller to cooperatively solve the collision avoidance problem for a set of vehicles. The method, called proximal message passing, is completely decentralized and needs no global coordination other than synchronizing iterations, allowing us to solve the avoidance problem extremely efficiently and in parallel. (Published in 2015 American Control Conference, Chicago, USA, July 2015)

Distributed Deep Q-Learning (paper, slides, code)

H. Y. Ong, K. Chavez, and A. Hong

  • We propose a distributed deep learning model to learn control policies directly from high-dimensional sensory input using reinforcement learning (RL). We adapt the DistBelief software framework to efficiently train the deep RL agents using the Apache Spark cluster computing framework.

Value Function Approximation via Low Rank Models (paper, slides, code)

H. Y. Ong

  • We propose a novel value function approximation technique for Markov decision processes that compactly represents the state-action value function using a low-rank and sparse matrix model. Under minimal assumptions, this decomposition is a Robust Principal Component Analysis problem that can be solved exactly via the Principal Component Pursuit convex optimization problem.


Player Behavior and Optimal Team Composition for Online Multiplayer Games (paper, poster, code)

H. Y. Ong, S. Deolalikar and M. Peng

  • The goal is to determine a set of descriptive gameplay style groupings and learn a predictor for win/loss outcomes. We present a machine learning framework for clustering player behavior and learning the optimal team composition for multiplayer online games.

Type II Diabetes Prediction with Incomplete Patient Record (manuscript, poster)

H. Y. Ong, D. Wang and X. S. Mu

  • We develop a smart Type II diabetes predictor that prompts high-risk patients to obtain diabetes testing given their electronic medical record. Our algorithm learns a Bayesian network and applies probabilistic inference to predict diabetes for patients with incomplete patient records.

Cooperative Vehicle Motion Planning with Collision Avoidance Constraints (manuscript)

H. Y. Ong and S. Deolalikar

  • We consider solving the multi-vehicle collision avoidance problem with a mixed integer program-based model predictive controller. A key advantage of our approach is the fidelity of our vehicle model, which incorporates nonlinear tire dynamics and stability constraints.

Truss Topology Design via Alternating Convex Optimization (manuscript, poster)

H. Y. Ong and C. Stansbury

  • We present a heuristic method of improving solutions to the truss topology design problem. Classically, convex optimization is used to size members subject to fixed attachment point locations. In our method attachment point locations are allowed to iteratively migrate small distances, resulting in an algorithm that intuitively learns appropriate shapes for bridges and buildings.