Cornell Tech Lightning Talks

Valts Blukis, Ilan Mandel, David Goedicke, Natalie Friedman, Travers Rhodes, PhD students, Cornell Tech


Location: Zoom

Time: 2:45p.m.


Natalie Friedman: Within human-robot interaction, I study how robots should be designed to move and behave in various contexts, based on the perception of social appropriateness.

David Goedicke: Our lab works mostly on resining implicit interaction for devices that to some degree make their own decision. I will show past work on Autonomous Vehicles (as large robots one sits in). How we use Virtual Reality to explore interaction, and my new research direction, which is Acoustically aware robots.

Valts Blukis: We study representation learning approaches for building robots that understand natural language in context of raw visual and sensory observations. I’ll present our recent work on mapping raw images and navigation instructions to physical quadcopter control, using a neural network model trained using simulated and real data. The model reasons about the need to explore the environment and incorporates geometric computation to predict which locations in the environment to visit. Finally, I’ll talk about the challenges when scaling representation learning methods to reason about previously unseen objects and environments.

Travers Rhodes: Variational Auto-Encoders (VAEs) have been known to “ignore” some latent-variable dimensions in their representations. This talk explores known results for what those representations look like for simplified, linear VAEs and presents some directions for future work on more complicated VAEs.

Addressing Spatial Data Challenges in Winegrape Vineyards

Dr. Nick Dokoozlian, E&J Gallo Winery; Dr. Hunter Adams, Cornell University;  Jonathan Jaramillo, PhD student, Cornell University.


Location: Zoom

Time: 2:45p.m.

Abstract:Vineyard blocks often contain a high degree of spatial variability in their vine growth, yield, and fruit composition.  This seminar will discuss technological solutions to measuring, modelling, and managing that spatial variability with the goal of reducing vineyard inputs and improving economic returns.

Safe Robot Motion Planning via Reachability Analysis

Shreyas Kousik, University of Michigan


Location: Zoom

Time: 2:55p.m.

Abstract: For robots such as autonomous cars, delivery drones, and manipulator arms, to interact with and around humans, it is critical to certify the safety of their motion planning algorithms. Motion planning typically uses a three-tiered hierarchy; first, the robot plans a path to achieve a goal, ignoring dynamics; then, it plans a trajectory subject to dynamics that attempts to follow the path; and finally, it tracks the trajectory with a controller. Methods exist that attempt to certify safety at each of these tiers, but often suffer a tradeoff between safety and performance. This talk presents a recent approach, called Reachability-based Trajectory Design (RTD), that leverages reachability analysis and optimization to enforce safety at the trajectory planning tier. RTD enables strict guarantees without sacrificing performance. The method is applied to cars, drones, and arms, to show its capability and versatility.

Soil and Soil Robotics: Applications and Challenges

Cross-listed from VIEN 4940,  Robert Shepherd and Anand Mishra, Cornell University


Location: Zoom

Time: 2:55p.m.

Abstract: Soil, the microbiome, and plant roots represent a critical frontier in terrestrial carbon sequestration. But the opacity of land, coupled with the heterogeneity and dynamic nature of the soil-microbe-root system, has severely limited direct observations. Here, we will discuss how to develop a suite of technology to provide dynamic access to the rhizosphere (root-soil interface) for multi-modal sensing of chemical and biological parameters with a focus on in-situ soil, water, and carbon interactions. In particular, we will discuss, nature of soil, plant, rhizosphere, and borrowing mechanism of biological organisms, especially plants. We will also see how plant root could be an ideal model for soil exploration robot and its mechanism of root growth, energy efficiency, sensing, and low fiction strategy. Also, we will have a look at our current robotic prototype (SoilBot), which aims to study soil properties, track the root inside the soil and study root excretions and microbiome exudation (metabolites and volatiles) in real-time.

Student meeting with David Fridovich-Keil

David Fridovich-Keil, University of California, Berkeley


Location: Zoom

Time: 3:30p.m.

Abstract:Autonomous systems are becoming more and more prevalent; meanwhile, both industry and academia are pushing the fundamental limits of what these systems can do. As autonomous systems become more sophisticated and tightly coupled to human society, their safety and effectiveness rely increasingly upon accurate modeling and prediction of human behavior. This talk will touch on work at multiple levels of abstraction, including low-level robust control and higher-level confidence-aware human prediction and motion planning. We will focus, however, on a new solution strategy for solving nonlinear N-player general-sum differential games, which arise in many motion planning problems of practical interest. The proposed algorithm is easily real-time, even for moderately large problems. We will discuss the core intuition behind the proposed approach, present convergence and complexity results, and demonstrate it through several examples.

Planning and Control of Multi-Agent Systems under Spatio-Temporal Logic Tasks

Lars Lindemann, KTH


Location: Upson 106 Conference Room Next to the Lounge

Time: 2:45p.m.

Abstract: Motivated by the recent interest in the distributed and provably-correct control of interconnected autonomous systems, this talk is about planning and control of dynamically coupled multi-agent systems under signal temporal logic (STL). STL allows to express complex tasks that impose temporal and spatial requirements on the multi-agent system. In the first part of the talk, a fragment of STL is considered and we show that an STL task from this fragment can be encoded into a control barrier function. We consider two cases, namely multi-agent systems under global and local, possibly conflicting, tasks. We then derive decentralized control barrier function-based control laws that ensure task satisfaction in the former case. In the latter case, dealing with conflicting local tasks, least violating solutions are found by relaxing tasks in the spatial domain. In the second part of the talk, the full STL language is considered by combining the previously derived decentralized control laws with efficient automata-based planning methods.

Integrating Robots and Ecology in Pollen-Limited Crops


Location: Upson 106 Conference Room Next to the Lounge

Time: 2:55p.m.

Abstract: Although the majority of the human diet stems from staple grains, approximately 75% of agricultural crops needs some amount of pollination. Global reliance on crop pollination is expanding, yet farmers in many parts of the world suffer from increasingly unpredictable yields stemming from dwindling populations of wild pollinators and unsustainable losses of managed bees. We are launching a research project in which we aim to integrate robotic and bio-hybrid agents into the crop ecosystem, to enable observation, estimation, and optimization of yield in pollen-limited crops. The core research challenge is to enable scalable and robust coordination of ubiquitous swarms composed of intelligent entities with varying degrees of capability, controllability, and cost. The project will revolve around a scheduling framework that assimilates and dispatches information and tasks at different levels of granularity, to permit robust progress by diverse agents in the face of failures and dynamic environments; we are also examining autonomous robots (drones and rovers) for managed pollination.

IROS Paper Exchange


Location: Upson 106 Conference Room Next to the Lounge

Time: 2:45p.m.

This Robotics Seminar (over February break) will be our first installment of paper exchanges. We will be providing review templates and other resources to help make the paper exchange productive. While it is timed a week before the IROS deadlines, people are welcome to come with drafts at varying stages for any venue. We will be implementing a “token” exchange system, where participants can gain tokens for reviewing others’ papers or speaking at robotics seminar. Participants can then use these tokens to ask for reviews and feedback. If you don’t have a draft at the moment, but would be happy to help out and review, please come and help out deadline-crunching peers! We will discuss more details about the token system at this seminar.”


Academic Paper Writing Clinic: Principles and Practice

Guy Hoffman, Cornell University


Location: Upson 106 Conference Room Next to the Lounge

Time: 2:45p.m.

Abstract: How does one write a good academic paper? What makes some papers easier to read than others? Are there techniques that can easily be applied to improve your paper? How do you overcome “blank-page syndrome”? In this workshop, I will share some of the lessons I have learned over years of writing academic and non-academic texts. I will analyze published papers and, if there is interest, propose strategies for students’ existing papers-in-process. Please send examples your own writing that you would like us to discuss at least 48 hours before the seminar.

Teaser: Here are two of Donella Meadows’s [] tips for writing an op-ed column:

1. Be clear, not fancy: Use everyday language. Be specific, not abstract.  Offer easily imaginable examples. Be sure your words make pictures in people’s heads. Be sure the pictures are the ones you intend.

2. Use most of your column for evidence: Tell stories, give statistics, show the impact of the problem or the solution on the real world. People can form their own conclusions if you give them the evidence. Don’t take much space for grand, abstract conclusions; let the reader form the conclusions.

Interactive Natural Language-based Person Search & Dynamics of Solid-Liquid Structures in Soft Robotics

Vikram Shree & Yoav Matia, Cornell University


Location: Upson 106 Conference Room Next to the Lounge

Time: 2:45p.m.


Vikram Shree: Interactive Natural Language-based Person Search

Today robots are equipped with rich sensors to enable different forms of interactions. Among these, using visual and natural language information is of particular interest and is commonly viewed as the most user-friendly way because of its frequent use in how humans interact with each other. One such problem that entails multi-modality of data is the task to find a person of interest (POI) in a crowd, based on natural language description about their appearance. In the talk, I will present my work on designing algorithms to systematically retrieve descriptions from a user about the POI.

Yoav Matia: Dynamics of Solid-Liquid Structures in Soft Robotics

In this work, we analyze the transient dynamics of solid-fluid composite structures. This is an interdisciplinary research subject, which lies on the border between theoretical fluid mechanics, soft-robotics and composite-structures. We focus on an elastic beam embedded with fluid-filled cavities as a representing common configuration. Beam deformation both creates and is induced by internal viscous flow, where changes to cavities’ volume are balanced by a change in axial flux. As a result, pressure gradients develop in the fluid in order to conserve mass, and stresses are induced at the solid-fluid interface; these in turn, create local moments and normal forces, deforming the surrounding solid and vice versa.

The results of the presented research can be applied to define the required geometric and physical properties of solid-fluid structures in order to achieve specific responses to external excitations, thus allowing to leverage viscous-elastic dynamics to create novel soft-actuators and solid-fluid composite materials with unconventional mechanical properties.