Rick's Home PageAbout MeResearch InterestsResearch ProjectsPublicationsWhy Do Research?TeachingResourcesPeopleLinks Of InterestContact Me

Research Projects

As a researcher, I am always involved in projects trying to push the bounds of modern computer science and mathematics in order to discover the next frontier. Below is a list of all the research projects in which I have participated. Since most people prefer to see the more recent works, my most recent projects are listed first followed by my past research projects.

Current Research

Below is a list of projects in which I am currently involved. Following each name is a brief description of the project as well as a list of other people involved. Related materials such as publications and posters will be listed with the associated projects as well.

Generalized Plan and Activity Recognition Models

Many plan, activity, and intent recognition (PAIR) models have been developed for specific problem domains such as recommender systems and personal desktop assistants. While these models perform extremely well for thier specific problem, they are not necessarily designed to work in other PAIR tasks outside the prescribed domain. We focus on developing integrated plan and activity recognition models applicable to a variety of domains.

Related Publications:
Freedman, Richard G. and Zilberstein, Shlomo. "Does the Human's Representation Matter for Unsupervised Activity Recognition?" The 2017 AAAI Fall Symposium Series: Artificial Intelligence and Human-Robot Interaction, p. 94-98, 2017. [pdf]
Freedman, Richard G. and Zilberstein, Shlomo. "Using Metadata to Automate Interpretations of Unsupervised Learning-Derived Clusters." Proceedings of the First Workshop on Human is More Than a Labeler, 2016. [pdf]
Freedman, Richard G.; Jung, Hee-Tae; and Zilberstein, Shlomo. "Temporal and Object Relations in Unsupervised Plan and Activity Recognition." Artificial Intelligence and Human-Robot Interaction: Papers from the AAAI Fall Symposium, p. 51-59, 2015. [pdf]
Freedman, Richard G. and Zilberstein, Shlomo. "Automated Interpretations of Unsupervised Learning-Derived Clusters for Activity Recognition." Workshop on Learning for Human-Robot Collaboration, 2015. [pdf] (updated version presented at the BeyondLabeler 2016 Workshop paper)
Freedman, Richard G.; Jung, Hee-Tae; and Zilberstein, Shlomo. "Temporal and Object Relations in Plan and Activity Recognition for Robots Using Topic Models." Artificial Intelligence and Human-Robot Interaction: Papers from the AAAI Fall Symposium (Technical Report FS-14-01), p. 70-72, 2014. [pdf]
Freedman, Richard G.; Jung, Hee-Tae; Grupen, Roderic A.; and Zilberstein, Shlomo. "How Robots Can Recognize Activities and Plans Using Topic Models." Artificial Intelligence and Robotics (AIRob): Papers Presented at the Twenty-Eighth AAAI Conference on Artificial Intelligence (Technical Report WS-14-01), p. 22-28, 2014. [pdf (extended version of the ICAPS 2014 Paper)]
Freedman, Richard G.; Jung, Hee-Tae; and Zilberstein, Shlomo. "Plan and Activity Recognition from a Topic Modeling Perspective." Proceedings of the Twenty-Fourth International Conference of Automated Planning and Scheduling (ICAPS), p. 360-364, 2014. [pdf]

Integrated Planning and Plan Recognition

We are extending a previously established transformation from a plan recognition problem over STRIPS domains to a classical planning problem (Ramírez and Geffner, 2010). In particular, we are adapting the method to real-time applications and integrating the process with planning for a responsive interactive system.

Related Publications:
Freedman, Richard G.; Fung, Yi Ren; Ganchin, Roman; and Zilberstein, Shlomo. "Responsive Planning and Recognition for Closed-Loop Interaction." The 2019 AAAI Fall Symposium Series: Artificial Intelligence and Human-Robot Interaction, p.1-5, 2019. [peer-reviewed pdf on arXiv (extended version of the ICAPS 2019 Demo Abstract)]
Freedman, Richard G.; Fung, Yi Ren; Ganchin, Roman; and Zilberstein, Shlomo. "Responsive Planning and Recognition for Closed-Loop Interaction." Demonstration Program for the Twenty-Ninth International Conference on Automated Planning and Scheduling, p.1-2, 2019. [pdf]
Freedman, Richard G. and Zilberstein, Shlomo. "A Unifying Perspective of Plan, Activity, and Intent Recognition." Proceedings of the Workshop on Plan, Activity, and Intent Recognition, p.1-8, 2019. [pdf]
Freedman, Richard G.; Fung, Yi Ren; Ganchin, Roman; and Zilberstein, Shlomo. "Towards Quicker Probabilistic Recognition with Multiple Goal Heuristic Search." Proceedings of the Workshop on Plan, Activity, and Intent Recognition, p.1-6, 2018. [pdf]
Freedman, Richard G. and Zilberstein, Shlomo. "Roles that Plan, Activity, and Intent Recognition with Planning Can Play in Games." Proceedings of the Workshop on Knowledge Extraction from Games, p.1-4, 2018. [pdf] [talk]
Freedman, Richard G. and Zilberstein, Shlomo. "Integration of Planning with Recognition for Responsive Interaction Using Classical Planners." Proceedings of the Thirty-first AAAI Conference on Artificial Intelligence, p.4581-4588, 2017. [pdf]
Freedman, Richard G. and Fukunaga, Alex. "Integration of Planning with Plan Recognition Using Classical Planners (Extended Abstract)." Artificial Intelligence and Human-Robot Interaction: Papers from the AAAI Fall Symposium, p. 48-50, 2015. [pdf]


Past Research

Below is a list of projects in which I was involved at some point. These entries may no longer be up to date.

Learning Physical Therapy by Demonstration

We are developing a framework for teletherapy where robots learn from physical therapists by demonstration. This will allow remote therapy to be performed with less frequent therapist supervision so that patients receive more care.

Related Publications:
Jung, Hee-Tae; Freedman, Richard G.; Takahashi, Takeshi; Wong, Jay-Ming; Zilberstein, Shlomo; Grupen, Roderic A.; and Choe, Yu-Kyong. "Adaptive Therapy Strategies: Efficacy and Learning Framework." Proceedings of the IEEE/RAS-EMBS International Conference on Rehabilitation Robotics, p.950-955, 2015. [pdf]
Jung, Hee-Tae; Freedman, Richard G.; Foster, Tammie; Choe, Yu-Kyong; Zilberstein, Shlomo; and Grupen, Roderic A. "Learning Therapy Strategies from Demonstration Using Latent Dirichlet Allocation." Proceedings of the Twentieth International Conference on Intelligent User Interfaces, p. 432-436, 2015. [pdf]

Restating the Collatz Conjecture

In mathematics, there are many problems that remain unsolved regardless of how simple they look. One example of such an unsolved problem is the Collatz Conjecture, also called the 3x+1 Problem amongst other names. The simple question is whether or not the following process will always generate a sequence of integers that contains 1 within a finite number of steps:
1. Select any positive integer x_0
2. Compute x_n = 3 * x_(n - 1) + 1 if x_(n - 1) is odd and x_n = x_(n - 1) / 2 if x_(n - 1) is even
3. Repeat step 2. using the newly obtained integer
We hope to use properties of the integer sequences generated by the function to find equivalent problems that help explain the orbits and why this problem is so difficult.

Related Publications:
Freedman, Richard G. and Robinson, Stephen B. "A Restatement of the Collatz Conjecture with Insights into Its Orbits Using a New Discrete Dynamical System." To be published in Pi Mu Epsilon Journal. [pdf coming soon!]
Related Posters:
Freedman, Richard G.; Robinson, Stephen B.; Rouse, Jeremy A. "Novel Approaches for Solving the Collatz Conjecture Using A New Coding Process." Wake Forest University's 5th Annual Research Day, 2011.
Freedman, Richard G. and Robinson, Stephen B. "Understanding Hailstone Sequences Using a New Coding Process." 2011 Joint Mathematics Meetings.

Using Natural Language Generation to Develop Assistive Technologies

Using natural language generation and other artificial intelligence techniques, we are developing assistive technologies for mobile devices to facilitate communication tasks for people with disabilities.

Mentors: Dr. Victor Paúl Pauca and Dr. William Turkett
Collaborators: Jingyi Guo
Related Publications:
Freedman, Richard G.; Guo, Jingyi; Turkett, William H.; Pauca, V. Paúl. "Hierarchical Modeling to Facilitate Personalized Word Prediction for Dialogue." Plan, Activity, and Intent Recognition: Papers from the AAAI 2013 Workshop, p.2-9, 2013. [pdf with fixed figures (some edges were missing in the models)]

Improving MAP Estimation Using Linear Programming and Quadratic Programming Relaxation

We are developing a new Maximum-a-Posteriori (MAP) estimation algorithm to further reduce/remove the error from current approximation methods. MAP is a method for learning the assignments of variables in a Markov Random Field (MRF) by finding the assignment that maximizes the likelihood of the MRF.

Related Publications:
Kumar, Akshat; Zilberstein, Shlomo; Toussaint, Marc. "Message Passing Algorithms for MAP Estimation Using DC Programming." Proceedings of the International Conference on Artificial Intelligence and Statistics, p. 656-664, 2012. [pdf]

Anytime Lifted Belief Propagation (ALBP)

Belief propagation is an inference technique that, for a factor graph representation of a world with evidence, can determine the marginal likelihoods of each variable's value assignments. However, for such domains as decision making, one only needs to know the marginal likelihoods for the assignment of a subset of the variables in the factor graph. ALBP accomplishes this by combining theorem proving techniques with lifted inference. I developed the first implementation of ALBP which we have used to empirically test the algorithm.

Mentors: Dr. Sriraam Natarajan and Dr. Rodrigo de Salvo Braz
Related Publications:
Freedman, Richard G.; de Salvo Braz, Rodrigo; Bui, Hung; Natarajan, Sriraam. "Initial Empirical Evaluation of Anytime Lifted Belief Propagation." International Workshop on Statistical Relational AI, 2012. [pdf]
de Salvo Braz, Rodrigo; Natarajan, Sriraam; Bui, Hung; Shavlik, Jude; Russell, Stuart. "Anytime Lifted Belief Propagation." International Workshop in SRL, 2009. [pdf]

Improving Error Correction in Compilers Using Natural Language Processing Techniques

Compilers translate human-written program code into machine-executable programs. However, it cannot complete the translation when there are errors in the code. We proposed a method employing natural language techniques to make the compiler develop a more semantic interpretation of the code. The intent was to improve error correction during the compilation process.

Computer Science Undergraduate Honors Thesis Advisor: Dr. Todd Torgersen
Related Papers:
Freedman, Richard G. Improving Error Correction in Compilers Using Natural Language Processing Techniques. Senior undergraduate honors thesis in computer science. Presented May, 2012. [pdf unavailable at the moment, sorry]

Measuring the SRB Entropy of Variations of the Expanding Map on the Circle

The expanding map on the circle may be visualized as follows: wrap an elastic band around a circle whose circumference is the length of the unstretched band, mark a point on the elastic band, unwrap the band and stretch it by some factor k > 1, rewrap the stretched band around the circle, and note where the marked point is now located on the circle. This process is repeated indefinitely with the next mark at the location of the point after rewrapping the band in the previous iteration. The change in the distribution of all points along the circle is a chaotic discrete dynamical system, and we use entropy to measure how chaotic these systems are from different observed perspectives. The SRB entropy measures chaos through the physical measure which is how we view it in the real world. Due to the intractable computations involved, we use numerical methods to approximate the entropy of the expanding map on the circle when k=2 and we do not uniformly stretch the band (that is, some parts are stretched more and others are strectched less). This work has generated some hypotheses regarding how the variations in the non-uniform stretching affect the SRB entropy.

Mathematics Undergraduate Honors Thesis Advisor: Dr. Miaohua Jiang
Related Papers:
Freedman, Richard G. On Variations of the SRB Entropy of the Expanding Map on the Circle. Senior undergraduate honors thesis in mathematics. Presented May, 2012. [pdf]

Rogo

Some students learn better from hands-on experience with a physical component. Since most programming courses teach languages such as Java, Python, or C/C++, the physical component of the students' hands-on experience is missing. To bring the results of their programs into the real world, we are developing Rogo, an introductory programming language for the iRobot Create. Inspired by such educational programming languages as Logo the Turtle and Carol the Robot, Rogo will provide very simple commands for controlling the Create including navigation, sound, light, and sensor reading. Students will be able to create more complex functions by combining the commands which will also introduce them to computational thinking (for example, three left turns is one right turn).

Mentors: Dr. William Turkett
Related Events: Robotics Demo Day (photographs may be viewed at WFU's CS Google+ page or at WFU's CS Facebook page)

Cardio View

CardioView is the name of a mobile app informing students about the availability of equipment in the on-campus gym at Wake Forest University. It was created by Rachel Black (now at Booz Allen Hamilton) and me for an independent study course under Dr. Daniel Cañas where we each focused on a different component. Rachel developed the front-end application and interface for the iPhone and continued this component after the semester for her senior honors project. I worked on the back-end reading data from sensors placed on the equipment and using the data to learn a simple classifier for whether or not the corresponding gym equipment was in use.

Project partners: Rachel Black
Mentors: Dr. Daniel Cañas
Related Posters:
Black, Rachel A.; Freedman, Richard G.; Cañas, Daniel A. "Cardio View: Equipment Use in the Miller Center." Wake Forest University's 5th Annual Research Day, 2011.
Related Papers:
Black, Rachel A. "Cardio View." Senior undergraduate honors project in computer science. Presented May, 2011. [pdf]

Multiagent Random Walks along a Torus

Cyber attacks on networks are dangerous since the malicious software can transport between servers and spread the damage. Thus it is important to locate malicious activity as soon as possible in order to avoid further harm. Malware checks on individual servers are computationally intensive and take a long time which makes them less ideal. A new nature-inspired approach to handling malware search over networks is a swarm multi-agent system called digital ants. I helped with some simple analyses of the effectiveness and efficiency of the digital ants along a torus-shaped network topology.

Mentors: Dr. Errin Fulp

The Information on Communication and Behavioral Study (ICBS)

ICBS was a game theoretic study in which human subjects were given short weekly quizzes based on a provided website link. Participants were sent a link to one of two websites containing information about current trends. The goal was to determine how people's behavior was influenced based on the information they were provided. I developed the sign-up website and a system to autonomously carry out the study.

Mentors: Dr. Fred Chen
Related Papers:
Chen, Frederick; Griffith, Amanda; Cottrell, Allin; Wong, Yue-Ling. "Behavioral Responses to Epidemics in an Online Experiment: Using Virtual Diseases to Study Human Behavior." PLOS One. Volume 8, Issue 1, 2013. e52814 [paper in html]



Rick's Home PageAbout MeResearch InterestsResearch ProjectsPublicationsWhy Do Research?TeachingResourcesPeopleLinks Of InterestContact Me