Jorge Guerra, '14

Student

Jorge Guerra, '14

Files

Cohort

2014

Biography

Jorge Guerra was born in Maracaibo, Venezuela and moved to St. Petersburg, Florida at the age of twelve. He is pursuing an undergraduate degree in Computer Engineering with a minor in Mathematics. In addition to being a Mcnair Scholar, Jorge is also an iCubed and Excel scholar. He has held various leadership positions in The Society of Hispanic Professional Engineers (SHPE), serving as the organization's president during his sophomore year. He was also a mentor for the Internet Science and Technology Fair (ISFT) program. Jorge has participated in different industrial internships. He has worked at Intel Co. as a Probe Product Engineer as well as a Product Development Engineer. Jorge has also worked for Boeing Co. as a Knowledge Based System Engineer. He conducted research during the summer of 2014 at University of California-Berkeley in the SUPERB-ITS program. Jorge currently works in the Intelligent System Lab (ISL) at the University of Central Florida as an undergraduate researcher. His research interests are in the areas of multi-agent systems and robotics applications.

Faculty Mentor

Avelino Gonzalez

Undergraduate Major

Computer Engineering

Future Plans

Ph.D. in Computer Science

Research

A Multi-Agent Collaboration System using Context Based Reasoning Conducted at the University of Central Florida as part of the ICubed National Science Foundation funded research project. Mentor: Avelino Gonzalez, Professor, Department of Electrical Engineering and Computer Science In this research, we aim to model the behaviors of virtual soccer players while they create complex offensive and defensive collaborative plays. We model these collaborative behaviors using Joint Intention Theory in the Context-Based Reasoning (CxBR) paradigm. This paradigm has had great success on modeling, understanding and creating behaviors functions based on team reasoning and collaborative behaviors. Our hypothesis is that implicit communicating contexts or current player's situations, allow for a better representation of collaborative behaviors than situational awareness. Using the structure of CxBR: Mission, Major Contexts and Sub-Contexts; a system process and take into consideration a set of sentinel rules that transition or activate a context which in turn possesses the actions an agent can do. This system is able to guide the player to accomplish a long term goal in a soccer domain. The soccer domain is a 2D virtual environment called Simple Soccer developed by Mat Buckland. Another goal for this research is to expand on the Context-Based Reasoning paradigm not only to represent collaborative behaviors among virtual agents but then study and implement it in real world applications.

Summer Research

A First-Order Open-Universe Partially Observable Markov Decision Process BLOG Model Conducted at the University of California-Berkeley for SUPERB-ITS. Mentor: Siddharth Srivastava, Ph.D., Electrical Engineering and Computer Science, College of Engineering, University of California-Berkeley Abstract: Through the used of an extended Probabilistic Programming Languages (PPL), we were able to come up with a probabilistic model to represent non-trivial descriptions of agent's ability for observation and action using the First Order Open University Partially Observable Markov Decision Process algorithm . This model can be use to motive the question of implementing Open-Universe POMDP decision planning process as a robot's main choice for sequential action decisions. We developed a probabilistic model that guided a robot to accomplish his goal of picking up an specific object. Using a Probabilistic programming languages (PPLs) such as Bayesian Logic (BLOG) provided us with a very useful method of designing and applying sequential planning models that deal with a lot of uncertainty in the world; more specifically such programming language was able to handle open-universe problems – the problems where exists uncertainty about the number of objects in the world and uncertainty about the identities of those objects – very efficiently. Also, implementing the POMDP algorithm, a decision-theoretic planning model for an agent, provided us with the tools needed for the agent to deal with a partially observable environment through stochastic actions. After a model was created, we ran a derived Open University Point-Based Value Iteration (OU PBVI) engine in order to process the model and produce the corresponding sequence of actions.

Summer Research Institution

University of California Berkeley

Disciplines

Computer Engineering | Computer Sciences

Jorge Guerra, '14

Share

COinS