profile

Julio Godoy

Ph. D
Department of Computer Science and Engineering
Universidad de Concepcion
Edmundo Larenas 219
Oficina 317
Concepcion, Chile
Office: 56-412203574
Email: juliogodoy at gmail dot com

I am an assistant professor at the Department of Computer Science and Engineering at Universidad de Concepcion. I obtained my Ph.D. at the University of Minnesota, working under the guidance of Dr. Maria Gini and Dr. Stephen J. Guy. My thesis work focused on machine learning algorithms for multi-agent navigation, where agents need to move in congested environments while avoiding collisions. My work aims at increasing the global efficiency of the agents' motions, using decentralized methods where agents have only local information. I am interested in reinforcement learning, multi-agent coordination algorithms.

I obtained a BSc in Engineering Science in Universided de Concepcion, Chile, where I also obtained a MSc in Computer Science. Since 2010, I have been pursuing my doctoral degree at the University of Minnesota.


Here you can find links to some of the projects I have worked on:

  • ALAN: Adaptive Learning for Multi-Agent Navigation.
  • PHOP: Progressive Hindsight Optimization.
  • C-Nav: Implicit Coordination for Multi-Agent Navigation.
  • VelPlan and SocialPlan: Safe and Efficient Navigation Among Heterogeneous Agents.

Multi-robot Task Allocation with Temporal Constraints

mrtw

Given a finite set of robots R, a finite set of tasks T. A task is defined by a location, assume (x,y), a duration, a time window defined by an early start time (EST) and latest finishing time (LFT) [EST,LFT]. The problem compute optimal or approximate allocations w.r.t to some cost measure (e.g makespan). We consider cases when tasks are known upfront, when tasks arrive dynamically, and navigational uncertainties. We propose a mixed integer linear (MILP) model, a decentralized auction model. Currently, we are developing algorithms for allocation with navigational uncertainties. AAAI Source: [Unavailable]. AAMAS Source: [pdf] (53.9kB)


Opponent Modelling in Transportation Networks

cop-drivers   Given a road network, a travel time function on the network and statistics about where speeding has occurred at given observation times, we find the allocation of police cars (simulated as autonomous agents) to roads so to minimize the number of speeding insidents. We provide a stochastic model for the problem in which states are the locations of agents, actions for police are to move or stay, and the action for driver agents are to stay within the speed limit, or exceed the speed limit by some amount. We offer an Weighted-Attraction-based learner and compare its performance against a reinforcement learner, and a random decision maker. Source: [pdf]


Short-term Predictions on Road Networks

cop-drivers   We propose a novel framework for predicting the paths of vehicles that move on a road network. The framework leverages global and local patterns in spatio-temporal data. From a large corpus of GPS trajectories, we predict the subsequent path of an in-progress vehicle trajectory using only spatio-temporal features from the data. Our framework consists of three components: (1) a component that abstracts GPS location data into a graph at the neighborhood or street level, (2) a component that generates policies obtained from the graph data, and (3) a component that predicts the subsequent path of an in-progress trajectory. Hierarchical clustering is used to construct the city graph, where the clusters facilitate a compact representation of the trajectory data to make processing large data sets tractable and efficient. Source: [HTML]


Search and Rescue Teamwork

cop-drivers   We propose coordination mechanisms for multiple heterogeneous physical agents that operate in city-scale disaster scenarios. Large scale disasters are characterized by limited and unreliable communications, dangerous events that may disable agents, uncertainty about the location, duration and type of tasks, and stringent temporal constraints on task completion times. In our approach, agents form teams with other agents that are in the same geographical area. Our algorithms either yield stable teams that are formed upfront and never change, or fluid teams where agents can change teams as need arises, or teams that restrict the types of agents that can belong to the same team. Our algorithms are tested using the RoboCup Rescue simulator. Source: [pdf]



Spring 2014

  • Introduction to Artificial Intelligence, CsCi 4511

  • Summer 2011

  • Structure of Computer Programming II, University of Minnesota, CsCi 1902

  • Spring 2011

  • Introduction to Artificial Intelligence, CsCi 4511
  • Please download my CV here: [PDF]

    I was born and grew up in the south of Angola. I speak Portuguese, Spanish, and English.
    I really love Math, and I continue to learn more of it. I also love running, playing soccer, cooking meat stews and hiking.
    I was very blessed to have travelled to many countries including Sweden, Swaziland, South Africa, Portugal, Spain and Italy. I enjoyed my time in all these countries.