2019 Computer Science Colloquium Series

Incorporating Real-World Semantics into Program Analysis of Robot Systems

John-Paul Ore

Wednesday, February 13, 2019
Centennial Engineering Center 1041
2:00-3:00 PM

Abstract:

Robotic software is plagued both by faults that menace all software (null-pointers, index-out-of-bounds) and also faults specific to its physical interaction with the real world, such as dimensional inconsistencies. These software hazards occur when developers incorrectly manipulate real-world quantities with physical units, such as confusing force with torque or measuring an angle in degrees instead of radians—something we have shown frequently happens in practice. We also found that existing solutions to these problems are time-consuming and error-prone. To address the state of the art, we designed a program analysis technique and its corresponding tool ’Phys’ to automatically detect dimensional inconsistencies in robotic software with minimal developer burden. Phys uses probabilistic reasoning and dataflow analysis to infer what variables mean in the real world. Phys works on systems that use the popular ‘Robot Operating System’ (ROS). I will present an evaluation showing that Phys has an 85% True Positive rate. I will show that dimensional inconsistencies lurk in at least 6% (211/3,484) of open-source robotic software repositories. I will further show the results of an empirical study showing that developers correctly identify the physical units of variables only 51% of the time, motivating our future work on automatically suggesting physical unit types. Finally, I will present a vision of future robotic software research enabled by our techniques that aims to help developers build robots with more reliable robotic software.

Bio:

John-Paul Ore is a Ph.D. candidate with the Computer Science and Engineering department at the University of Nebraska–Lincoln. His research is in software engineering and field robotics. His Ph.D. work focuses on how to automatically detect dimensional inconsistencies in robotic software without time-consuming developer annotations. Specifically, he builds techniques and tools that infer physical unit types (like ‘meters-per-second’) using probabilistic reasoning to combine facts from dataflow with evidence from uncertain sources like variable names. He also performs empirical studies of developers to assess their ability to make decisions about robotic software. Overall, his goal is to help robotic system developers create better and safer systems. John-Paul received an Othmer fellowship, a US Patent for Aerial Water Sampling (#US9606028B2), ‘Best Masters Thesis’ Award (2014), ‘Best Tool Demonstration’ (ISSTA’17), and is on the program committee for Robotic Software Engineering Workshop (RoSE, part of ICSE’19). He has a B.A. in Philosophy from the University of Chicago. More info and videos: https://cse.unl.edu/~jore

The Price of Ignorance in Swarm Robotic Central Place Foraging: An Analytical Perspective

Abhinav Aggarwal

Wednesday, February 6, 2019
Centennial Engineering Center 1041
2:00-3:00 PM

Abstract:

Renewed interest in Central Place Foraging has resulted from the need for robotic in-situ resource utilization solutions as a means to support habitation on other worlds. Practical algorithms that are provably scalable are needed. A key factor limiting the performance of foraging algorithms is the awareness of the bot(s) about the location of food items around the nest. From this perspective, a metric called Price of Ignorance is introduced. It measures how much time an ignorant bot takes relative to an omniscient forager for complete collection of food items in the arena.

In this talk, I will talk about this metric and use one deterministic (DASA) and one randomized foraging algorithm (BalCPFA) for a comparison study. These two algorithms are based on the top two algorithms from the recent Swarmathon competition. The analysis confirms the recent empirical claim by Lu et al. (ICRA 2019) that a deterministic spiral search can be expected to outperform a random walk based search (based on a Gazebo simulation). It also shows that even when the bots deploy perfect site fidelity, BalCPFA is unable to outperform DASA, in expectation. Furthermore, upon analyzing the effect of depletion of food items from the arena on the foraging efficiency, this analysis allows us to conclude that this effect is likely to be one of the key factors causing the BalCPFA lag behind DASA.

Bio:

Abhinav Aggarwal is a Ph.D. candidate at the University of New Mexico Computer Science Department, working with Prof. Jared Saia on robust interactive communication protocols and resource competitive analysis. He likes working on interesting mathematically challenging problems, mainly at the intersection of security and theoretical aspects of distributed computing. He has interned at several places like Microsoft, Google, VISA Research and Cornell University for his projects that span across different aspects of secure and fault-tolerant distributed systems. He has also served on various graduate student organizations on the campus, including CSGSA, GPSA, and ISA.

Complex and High-Dimensional Motion Planning Under Uncertain Conditions

Lydia Tapia, PhD

Wednesday, January 30, 2019
Centennial Engineering Center 1041
2:00-3:00 PM

Abstract:

Mankind is on the cusp of a robotics revolution. Soon, cars will drive themselves and our packages will be autonomously delivered by flying robot. Despite these advances, there is one aspect of navigation that robots are currently unable to handle well: uncertainty. Navigation uncertainty comes from many sources, both internal to the robot, e.g., control or localization uncertainty, or external to the robot, e.g., changes in or uncertainty of the world around the robot. In this talk, we will address multiple forms of uncertainty that impact autonomous navigation. First, we consider navigation in environments that are changing stochastically. Our methods are the first to directly integrate stochastic changes that occur during navigation and provide real-time capable solutions for navigation. Next, we consider transition uncertainty that occurs when an action is taken but the outcome is unexpected. Through adaptation of learned plans, we demonstrate adjustment to certain forms of transition uncertainty. Finally, we consider model uncertainty, a lack of precision or error in the world model used to navigate. While this form of uncertainty often causes unintended collisions, we demonstrate how to quantify and adjust to predicted collisions. Application of our methods spans both robotics and biological domains. In the robotics domain, we will demonstrate our solutions on autonomous vehicle navigation, aerial vehicles, and manipulation. In the biological domain, we investigate the impact of uncertainty in the simulation of antibody assembly, where multiple molecules are moving and interacting on a cell membrane.

Bio:

Lydia Tapia, PhD is an Associate Professor in the Department of Computer Science at The University of New Mexico. She received her Ph.D. in Computer Science from Texas A&M University and her B.S. in Computer Science from Tulane University. Her research contributions are focused on the development of computationally efficient algorithms for the simulation and analysis of high-dimensional motions for robots and molecules. Specifically, she explores problems in computational structural biology, motion under stochastic uncertainty, and reinforcement learning. Based on this work, she has been awarded two patents, one on a novel unmanned aerial vehicle design and another on a method to design allergen treatments. Lydia is the recipient of the 2016 Denice Denton Emerging Leader ABIE Award from the Anita Borg Institute, a 2016 NSF CAREER Award for her work on simulating molecular assembly, and the 2017 Computing Research Association Committee on the Status of Women in Computing Research (CRA-W) Borg Early Career Award.

The Tale of the Three Little Speakers and the Big Bad Wolf, or Why Do We Even Come to this Seminar?

Trilce Estrada, PhD

Wednesday, January 23, 2019
Centennial Engineering Center 1041
2:00-3:00 PM

Abstract:

Have you ever wonder why colloquium is a required class in the department of Computer Science? Have you ever asked yourself why (oh why) do you have to listen to a talk, or what do you even gain from attending to a research seminar? By now, research seminars are part of your academic life. They are designed to showcase a broad variety of topics or to bring different and fresh perspectives to the table. But depending on the culture of the department, the colloquium can be vibrant and exciting, or the most boring and pointless experience. In this meta-colloquium we will talk about the intrinsic vs. extrinsic motivators to attend to a research seminar, we will discuss strategies to formulate questions and engage with speakers, and we will interactively analyse a few case studies. Finally the goal of this talk is to set up foundations to maximize your gain, as a student, from this colloquium series.

Bio:

Trilce Estrada, PhD is an assistant professor in the department of Computer Science at the University of New Mexico. Her research interests include self-managed distributed systems, Big Data analysis, crowd sourcing, and machine learning. Recently, she was awarded the National Science Foundation's Early Career Award for the proposal entitled CAREER: Enabling Distributed and In-Situ Analysis for Multidimensional Structured Data.

Computational Design and Fabrication for All

Leah Buechley, PhD

Wednesday, January 16, 2019
Centennial Engineering Center 1041
2:00-3:00 PM

Abstract:

Computer Science for All (CS4All) is a new effort whose goal is to provide all K-12 students in the US with access to a CS education. Since it was announced in 2016, the initiative has gathered steam and school districts across the country are teaching their first computing classes. It is an exciting time for researchers in computer science and education; there is tremendous opportunity to shape the foundation of a new educational movement.

This talk will advocate for an approach to K-12 CS education that prioritizes young people's interests and engagement. I will argue that integrations of computing with design and hands-on making provide especially promising opportunities for deep engagement and learning in CS. I will survey relevant educational research, and present examples of how students from diverse backgrounds can create beautiful, meaningful artifacts by blending CS, design, and fabrication. I will present my own work in this area and discuss the exciting array of research opportunities presented by the intersection of CS4All with the emerging field of computational fabrication.

Bio:

Leah Buechley is a designer, engineer, and educator. Her work explores integrations of computing, electronics, and design. She has done foundational work in paper and fabric-based computing. Her inventions include the LilyPad Arduino, a construction kit for sew-able electronics. She currently runs a design firm, Rural / Digital, that explores playful integrations of technology and design. Previously, she was an associate professor at the MIT Media Lab, where she founded and directed the High-Low Tech group. Her research was the recipient of an NSF CAREER Award and the 2017 Edith Ackerman award for Interaction Design and Children. Her work has been exhibited internationally in venues including the Exploratorium, the Victoria and Albert Museum, and Ars Electronica and has been featured in publications including The New York Times, Boston Globe, and Wired. Leah received a PhD in computer science from the University of Colorado at Boulder and a BA in physics from Skidmore College.