TEA Lab: Trustworthy Engineered Autonomy

Safe Autonomous Systems @ University of Florida ECE

Recent News

  • Maxwell Ruyle Outstanding Student Leadership Award
    Maxwell wins Outstanding Student Leadership Award

    Maxwell Ruyle, a Mechanical Engineering undergraduate student, received the Bill “Roto” Reuter & Peter Nicholas Outstanding Student Leadership Award from the Department of Mechanical and Aerospace Engineering for leading activities in many contexts:

    • A founding member of Gator Autonomous Racing and a lead of the GAR mechanical team
    • Designer of the mechanical subsystem of the TEACar
    • A founder of a startup focused on defense innovations.
    • Industry chair of the UF ASME chapter

    Congratulations!

  • Five interacting themes of CPS resilience
    New preprint: a broad view of CPS resilience

    Ivan took part in a large many-university effort to summarize the state of resilient cyber-physical systems (CPS) and the outlook for future research in this area. Five themes have emerged.

    Citation:

    • Saurabh Bagchi*, Hyunseung Kim*, Tarek Abdelzaher, Homa Alemzadeh, Somali Chaterji, Glen Chou, Yuying Duan, Fanxin Kong, Michael Lemmon, Yin Li, Mengyu Liu, Wenhao Luo, Meiyi Ma, Sibin Mohan, Ayan Mukhopadhyay, Melkior Ornik, Dimitra Panagou, Kristin Yvonne Rozier, Ivan Ruchkin, Huajie Shao, Sze Zheng Yong, Majid Zamani, Xugui Zhou.
      Digital Guardians: The Past and The Future of Cyber-Physical Resilience [Arxiv]
      In submission, 2026. * Co-first authors.
  • Two people in suits standing in front of a poster called "Defensive Maneuvering Under Partial Observability in Autonomous Racing"
    Posters at Undergraduate Spring Symposium 2026

    Congratulations to Trevor and Chris (and Vignesh) on their presentations of undergraduate-led research projects!

    Featured above: Chris Oeltjen and Vignesh Saravanan with their defensive maneuvering work.

    Featured below: Trevor Turnquist with his calibrated filtering work.

  • Mathias wins ECE Undergrad Research Excellence Award

    Congratulations to Mathias Gast on his well-deserved research award! In the TEA Lab, he has been working on soundly abstracting continuous systems, resulting in multiple paper submissions.

    Looking forward to his upcoming contributions!

  • New preprint: latent-entropy anomaly detection

    In collaboration with ECE colleagues, we have put out a new variant of our unsupervised anomaly detection pipeline. This one uses a latent entropy loss to scramble the latent space, making anomalies harder to reconstruct (and hence easier to detect).

    No supervision (including a normal-only dataset) needed!

    Citation:

    • Yuang Geng, Junkai Zhou, Kang Yang, Pan He, Zhuoyang Zhou, Jose C. Principe, Joel Harley, Ivan Ruchkin.
      MLE-UVAD: Minimal Latent Entropy Autoencoder for Fully Unsupervised Video Anomaly Detection [Arxiv].
      Preprint, 2026.
  • Ivan presents LLM confidence calibration at Shonan

    Ivan had the honor of attending an invitation-only visionary workshop #235 on LLM-guided assurance and synthesis for CPS in Shonan, Japan. He presented the lab’s work on calibrating chain-of-thought confidence by discovering temporal patterns with Signal Temporal Logic.

    Materials:

    Some of the prominent debates at the workshop included:

    • What does the probability of LLM choices have to do with the probability of LLM mistakes?
    • Are world models necessary for intent?
    • Where do specifications for LLMs come from, and are they truly separate from data?
    • What is the equivalence class of semantically valid formalizations of natural language?
    • How to combine the perfection of formal methods and the magic of AI?
    • How is the explainability of LLM states different from the explainability of LLM outputs?
    • Should we prioritize syntactic or semantic robustness in reasoning?
    • How to establish multifaceted connections between the modalities of sensing (camera/lidar data), reasoning (language, both natural and formal), and control (actions)?
    • How do you expect the robot to clean dishes well if you did not teach it?
    • Does Lean have enough support for future, not-yet-existent mathematics?
    • What aspects of agent-based CPS engineering should we trust more, and which – less?

  • Ivan named Malachowsky Family Endowed Rising Star

    Big thanks to the Malachowsky Family for supporting our AI research! Also, congratulations to Alina and Yingying.

    Onwards!

    Links:

  • New preprint: verifiable deterministic world models

    Our exploration of world models for system assurance resulted in a semi-predictable but currently unfashionable choice: removing randomness and uncertainty from the latent space made world models more verifiable (although a tiny bit less picture-perfect). More surprisingly, this step made the behaviors produced by them more relevant to the real world.

    As a result, we were able to transfer guarantees obtained on verifiable deterministic world models to the real image-based system. Intriguing!

    Citation:

    • Yuang Geng, Zhuoyang Zhou, Zhongzheng Zhang, Siyuan Pan, Hoang-Dung Tran, Ivan Ruchkin.
      Deterministic World Models for Verification of Closed-loop Vision-based Systems [Arxiv] [Github].
      Preprint, 2025.
  • STL CoT confidence = most innovative poster

    Zhenjiang and Ani presented a poster with their work on chain-of-thought confidence with signal temporal logic at the Annual Nelms IoT Conference. The core idea of this research is to find patterns in LLM confidence that tend to correlate with correct and incorrect answers. Then, these patterns can be used to determine the confidence, i.e., the chance of the LLM providing the correct answer.

    Zhenjiang and Ani at their poster

    They were awarded the Most Innovative Poster Award by UF Innovate, which included a $500 cash prize. Congratulations!

    In the meantime, a couple of other fun events happened around that time:

    Richard Yang and Sriram Yerramsetty gave a demo of autonomous Roboracer overtaking at the Nelms Conference
    Gator Autonomous Racing hosted its first Build Day to start making a new crop of autonomous racing cars
  • Axis aligned
    New preprint: statistical-symbolic verification of perception

    Our collaboration with RPI has yielded an extended and improved version of our NeuS’25 paper: combining conformal prediction for neural perception with reachability analysis for the dynamics and control. This problem required constructing a discrete abstraction of the perception neural net, which we did with a genetic algorithm.

    Citation:

    • Yuang Geng*, Thomas Waite*, Trevor Turnquist, Radoslav Ivanov†, Ivan Ruchkin†.
      Statistical-Symbolic Verification of Perception-Based Autonomous Systems using State-Dependent Conformal Prediction. [Arxiv]
      In submission to the ACM Transactions on Embedded Computing Systems (TECS), 2025. * Co-first authors. † Co-last authors.
  • ECE showcases new club: Gator Autonomous Racing

    This semester marks a major development: the Gator Autonomous Racing (GAR) student club/design team was officially spun off from the TEA Lab.

    This week, the club has put together an impressive showcase with two racing cars (F1/tenth, aka RoboRacer) in the middle of Malachowsky Hall. The demonstration has attracted a lot of attention!

    A huge thanks to these guys for putting the showcase together:

  • New preprint: a survey of CPS assumptions

    We’ve put in a big effort to find, categorize, and analyze assumptions and guarantees in papers on cyber-physical systems since 2014. Now we’re happy to release the results!

    Citation:

    • Chengyu Li, Saleh Faghfoorian, Ivan Ruchkin.
      What Does It Take to Get Guarantees? Systematizing Assumptions in Cyber-Physical Systems [Arxiv].
      Preprint, 2025.

    We are also sharing our database of analyzed papers and assumptions.

    • Not seeing your favorite paper there? Fill out the form to let us know, and we’ll consider including it into the survey (subject to the inclusion criteria).
  • Ivan does publicity for a neuro-symbolic conference

    Ivan Ruchkin is serving as the publicity chair of the 3rd International Conference on Neuro-Symbolic Systems (NeuS) 2026.

    Looking forward to your submissions!

  • Jordan presents V&V for vision-based systems at ATVA

    Jordan Peper went all the way to Bengaluru, India, to present our work (in collaboration with UIUC) on unified verification and validation of vision-based autonomy at the International Symposium on Automated Technology for Verification and Analysis (ATVA). Allegedly, this is a hot problem, but the abstraction is quite complex. That’s what it takes — for now.

    Citation:

  • AutoGators win Most Innovative @ Autonomy Hackaton

    Congratulations to the team AutoGators (Krish Kapadia, Yilin Liu, Zhenjiang Mao, Ishaan Sen, Zhongzheng Zhang, Zhuoyang Zhou) on winning the “Most Innovative Solution” Award ($10K) at the Mission Autonomy Hackathon organized by AWS and Vanderbilt. 

    Here is the problem they solved: “Given a swarm of autonomous aerial drones tracking a resupply convoy, use aerial imagery to dynamically detect threats along the convoy’s path and reroute the convoy to avoid them with minimal delay.” 

    :

    Code:

  • IROS showcase: world models, image repair, data cleaning

    Ivan went all the way to Hangzhou, China, to present several research works on world models, image repair, and data cleaning.

    1. An invited talk “𝐑𝐞𝐥𝐢𝐚𝐛𝐥𝐞 𝐖𝐨𝐫𝐥𝐝 𝐌𝐨𝐝𝐞𝐥𝐬: 𝐏𝐡𝐲𝐬𝐢𝐜𝐚𝐥 𝐆𝐫𝐨𝐮𝐧𝐝𝐢𝐧𝐠 𝐚𝐧𝐝 𝐒𝐚𝐟𝐞𝐭𝐲 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧” at the Building Safe Robots: A Holistic Integrated View on Safety from Modelling, Control & Implementation Workshop (sponsored by NOKOV Motion Capture)
      – Featuring the work by Zhenjiang Mao, Mrinall Umasudhan, and Jordan Peper
    2. A paper presentation “𝐆𝐞𝐧𝐞𝐫𝐚𝐥𝐢𝐳𝐚𝐛𝐥𝐞 𝐈𝐦𝐚𝐠𝐞 𝐑𝐞𝐩𝐚𝐢𝐫 𝐟𝐨𝐫 𝐑𝐨𝐛𝐮𝐬𝐭 𝐕𝐢𝐬𝐮𝐚𝐥 𝐂𝐨𝐧𝐭𝐫𝐨𝐥”
      – Featuring the work of Carson Sobolewski, Zhenjiang Mao, and Kshitij Maruti Vejre
    3. A paper presentation “𝐔𝐧𝐬𝐮𝐩𝐞𝐫𝐯𝐢𝐬𝐞𝐝 𝐀𝐧𝐨𝐦𝐚𝐥𝐲 𝐃𝐞𝐭𝐞𝐜𝐭𝐢𝐨𝐧 𝐈𝐦𝐩𝐫𝐨𝐯𝐞𝐬 𝐈𝐦𝐢𝐭𝐚𝐭𝐢𝐨𝐧 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐟𝐨𝐫 𝐀𝐮𝐭𝐨𝐧𝐨𝐦𝐨𝐮𝐬 𝐑𝐚𝐜𝐢𝐧𝐠”
      Featuring the work of Yuang Geng, Yang Zhou, Yuyang Zhang, Zhongzheng Zhang, Kang Yang, Tyler Ruble, and Giancarlo Vidal

    Here are the paper citations on which these presentations were based:

  • Trevor and Jordan win student research awards at ESWEEK

    Congrats to Trevor Turnquist and Jordan Peper on winning the First Undergraduate and Runner-Up Graduate Awards at the ACM Student Research Competition hosted at the Embedded Systems Week 2025!

    Photos from the event:

    The papers relevant to these competition submissions:

  • Demos at HWCOE celebration and dean’s tailgate

    The TEA Lab and the newly formed Gator Autonomous Racing (GAR) club collaborated on two back-to-back demos:

    1. A camera-based DonkeyCar demo for the 10th year anniversary of the naming of UF engineering by Herbert Wertheim. This demo on October 3 was led by Zhongzheng Zhang and Tyler Ruble.
    2. A lidar-based RoboRacer demo for the Dean’s tailgate before the UF-UT Austin football game. This demo on October 4 was led by Ruben Gonzalez-Vera, Sriram Yerramsetty, Richard Yang, and Christopher Oeltjen.

    Congrats to the students on successfully demoing our autonomous racing technology!

  • New preprint: online friction estimation for racing

    Our lab pushed out an experimental project on detecting slip and estimating the tire friction from the onboard sensors (lidar & IMU) on RoboRacer (aka F1/10) cars. No fancy models, no sophisticated data collection, no need for post-processing. It turned out pretty accurate!

    • Christopher Oeltjen*, Carson Sobolewski*, Saleh Faghfoorian*, Lorant Domokos, Giancarlo Vidal, Ivan Ruchkin.
      Online Slip Detection and Friction Coefficient Estimation for Autonomous Racing [Arxiv] [Video].
      Preprint, 2025. * Co-first authors.

  • Ivan presents conservative perception abstractions at Allerton

    Ivan talked about conservative abstractions of perception-driven systems at the University of Illinois Urbana-Champaign in the Allerton Conference. The rumor is that these abstractions are too conservative.

    Citation:

  • New preprint: unified V&V for vision systems

    In collaboration with UIUC researchers, we have developed a methodology to build uncertainty-aware models (imprecise Markov decision processes) of vision-guided autonomous systems. These models offer a unified methodology for their verification (to get safety guarantees) and validation (to quantify the applicability of these guarantees to the real world). Accepted at ATVA 2025, this paper is a step in the long journey towards more practical formal methods for real-world autonomy.

    Citation:

  • Ivan talks about high-dimensional verification at USC

    The talk included methods for dealing with the high dimensionality of perception and state space. Setting a duration record for Ivan’s research talks, it lasted for 90 minutes (thanks to many insightful questions!).

  • New NSF project on verifiable safety under visual shifts

    We are excited to start the VISUALS project: Verifiable Information-Theoretic Safety Under Augmented Latent Shifts, in collaboration with Yuheng Bu (UCSB) and Jose Principe (UF), sponsored by the NSF EPCN program.

    This project aims to create an end-to-end methodology to model, analyze, quantify, detect, and adapt to changes in the visual environment of an autonomous system. It will bring together techniques and insights from formal methods, information theory, and uncertainty quantification.

  • New preprint: how safe will I be given what I saw?

    An extension of our modular family of learning-based safety predictors from L4DC 2024, now with transformers and quantization!

    Citation:

  • TEA Lab hosts incoming freshmen

    This summer, the TEA Lab welcomed a group of incoming UF freshmen as part of the STEPUP (the Successful Transition and Enhanced Preparation for Undergraduates program). The visiting students learned about the importance of safe and trustworthy autonomy, got to poke around the autonomous racing cars, and asked great questions!

  • Ivan talks about conformal reachability at CAV

    ​Ivan went all the way to Croatia to tell people how to put conformal prediction in a closed loop at the International Conference on Computer-Aided Verification (CAV). Doing so would let you verify autonomous systems with neural networks of any size (yes, even a VLA model like RT-2!). 

    The decisive question is, to apply conformal prediction at the perception level or at the control level? 

    1. If you apply conformal prediction at the perception level, you are making bounds on perception error. If so, it is wise to take into account how this error changes both over state and over time. Then state-based conformal prediction is at your service, per Ivan’s talk at the Third Workshop on Trustworthy Autonomous Cyber-Physical Systems (TACPS)​.

      Citation:
    2. If you apply conformal prediction at the control level, then you have a whole menu of options: single-step vs trajectory level, states vs actions. Turns out all these options lead to slightly different guarantees to bridge the gap between high dimensions (images) and low dimensions (states). These insights found their way into the poster that Ivan presented at the International Symposium on AI Verification (SAIV). 

      Citation:


    Photo credit: Taylor Johnson

  • Our world models are taking off

    Our recent dive into world models is blossoming in several intriguing directions: multimodality, hallucinations, and modular verification. While we’re pushing these directions forward, take a look at a nice overview article about our research on world models.

  • New NSF project on confidence calibration under anomalies

    The last decade has seen a flourishing of detection capabilities for various anomalies and out-of-distribution samples. However, the question of what an autonomous system should do after it detects an anomaly is incredibly challenging and seems to be nowhere near a satisfying answer. 

    A reasonable step after detecting an anomaly is to figure out, in as much detail as possible, how much it affects the operation of the system: (a) How much are the sensing, perception, planning, control, or the environment affected? (b) How much are systemic properties, like safety, affected? 

    This new NSF CPS project will develop a framework called Methodology for Anomalous Safety Confidence (MASC). This framework will adjust (on the fly) the system’s confidence in its own safety/correctness based on the anomalies that it is detecting. Looking forward to the exciting research ahead! 

    Links: 

  • New preprint: chain-of-thought confidence with STL

    We converted a SAS course project to a workshop paper about how to calibrate the confidence in chain-of-thought reasoning using a temporal logic formula: 

    Stay tuned for more!

  • Ivan presents principles of world modeling at NeuS 2025

    Ivan revisited his old grazing grounds in Philly to present 4 principles for making world models more physically grounded. There was an intense discussion of whether purely symbolic simulators should count as generative world models. 

    Citation: 

    In the meantime, the RPI collaborators Thomas and Rado presented a joint work on state-based conformal prediction.

    Citation: 

  • Zhongzheng and Yuyang give racing demo to Nelms family
    Thanks to Yuyang and Zhongzheng for impressing the visitors from the generous Nelms family on a new racing track! 
  • Jordan & Ivan present world models at ICRA 2025

    To the audience’s excitement, Jordan and Ivan presented the lab’s work on principles of physically interpretable work models in two capacities: 

    1. As a spotlight poster at the Workshop on Foundation Models and Neuro-Symbolic AI for Robotics (FMNS)
    2. As a late-breaking result poster in the main ICRA conference. 

    Citation: 

  • Jordan and Ivan present at CPS-IoT Week 2025
    Several events transpired at the CPS-IoT Week in Irvine, CA: 
    1. Jordan presented his poster (pictured) on probabilistic verification & validation at HSCC. 
    2. Ivan presented his collaborative work on imprecise neural networks at HSCC. 
    3. Ivan chaired the ICCPS poster/demo session and a couple of paper sessions, and also judged posters in the PhD forum. 
  • Lorant wins the student leader and best poster awards

    Congratulations to Lorant Domokos on winning two (!) awards from the Department of Mechanical and Aerospace Engineering (MAE) at the end of his senior year: 

    • 2025 MAE Student Leadership Award
    • 1st Place Undergraduate Poster at the 2025 MAE Poster Competition
      • (The poster on drift detection and friction estimation can be found here.)
  • Carson & Lorant present at the UF Spring Symposium
    Carson Sobolewski and Lorant Domokos presented their posters at the UF Spring Undergraduate Research Symposium 2025 as part of their scholarship programs: Allegedly, Chris Oeltjen was also in attendance. 
  • New preprint: conservative perception abstractions
    A new preprint is out on low-dimensional symbolic models of deep visual perception that enable conservative (i.e., non-overconfident) safety analysis.  Citation: 
    • Matthew Cleaveland, Pengyuan Lu, Oleg Sokolsky, Insup Lee, Ivan Ruchkin. Conservative Perception Models for Probabilistic Model Checking [Arxiv]. Preprint, 2025.
  • Ivan receives the NSF CAREER Award
    While everyone and their brother are chasing guarantees for autonomous systems, their assumptions are being overlooked. Ivan got the prestigious NSF CAREER grant to fix that problem: the new project will focus on the careful understanding, modeling, validation, and monitoring of important assumptions in learning-based autonomous systems.  More: 
  • New preprint: generalizable image repair
    Advanced GANs make short work of previously unseen image corruptions. Update: accepted to IROS 2025! Citation: 
    • Carson Sobolewski, Zhenjiang Mao, Kshitij Vejre, Ivan Ruchkin. Generalizable Image Repair for Robust Visual Autonomous Racing [Arxiv] [Poster] [Github] [Video]. Preprint, 2025.
  • New preprint: principles for interpretable world models

    Our new paper articulates four key principles for physical interpretability of world models. We paint a broader picture on neuro-symbolic world models, beyond our recent preprint on a specific technique for physically interpretable world models for trajectory prediction.  

    Update: accepted and presented at NeuS 2025! It also got publicized at ICRA.

    Citation: 

     

  • New preprint: state-based conformal prediction

    Our first collaborative paper on the NSF Neuro-Symbolic Bridge project with RPI is online! It develops a novel way to get tight conformal prediction bounds on perception error in order to improve the accuracy of reachability verification. 

    Update: published and presented at NeuS’25!

    Citation:

  • TEA lab does double racing demos for Spring Visit
    Great job to those who put together the demos for the ECE and MAE Spring Visits, particularly Zhongzheng and the F1/10 team!
  • Neurostrata
    New preprint: stratified neuro-symbolic architecture

    Check out our nice and short position paper. The key idea is to intermingle neural components and symbolic knowledge at each level of the autonomy stack.  

    Update: published in FSE’25!

    Citation:

  • Ivan co-chairs the poster/demo session at ICCPS 2025
    The International Conference on Cyber-Physical Systems (ICCPS) 2025 is seeking poster and demo submissions for its 16th iteration, in Irvine, CA. Details can be found here. 
  • Ivan co-chairs the poster/demo session at ICCPS 2025
    The International Conference on Cyber-Physical Systems (ICCPS) 2025 is seeking poster and demo submissions for its 16th iteration, in Irvine, CA. Details can be found here. 
  • New preprint: physically interpretable world models

    Our recent preprint develops an architecture and a training method to give latent states physical meaning in the context of trajectory prediction:

    • Zhenjiang Mao, Ivan Ruchkin.
      Towards Physically Interpretable World Models: Meaningful Weakly Supervised Representations for Visual Trajectory Prediction [Arxiv] [Poster].
      Preprint, 2024. 

     

  • MANY posters, demos, awards at NELMS IoT conference

    Congratulations to many students from TEA lab presenting their work and getting recognition!

    Demos

    Mrinall and Sam show off their DonkeyCar
    Mrinall and Sam show off their DonkeyCar
    Mrinall and Sam show off their DonkeyCar
    Mrinall and Sam show off their DonkeyCar
    Lorant, Carson, Chris, and Anthony show off their F1/10 car
    Lorant, Carson, Chris, and Anthony show off their F1/10 car
    Lorant, Carson, Chris, and Anthony show off their F1/10
    Lorant, Carson, Chris, and Anthony show off their F1/10 car

    Posters

    Sam and his work on language enhanced latent representations for out of distribution detection in autonomous driving
    Sam and his work on language-enhanced OOD detection
    Yuang and his work on bridging-dimensions confident reachability for high-dimensional controllers
    Yuang and his work on high-dimensional verification
    Zhenjiang and his work on interpretable world models
    Zhenjiang and his work on interpretable world models

    Awards

    First place for the F1/10 demo
    First place for the F1/10 demo
    First place for the F1/10 demo
    First place for the F1/10 demo
    First place for the F1/10 demo
    First place for the F1/10 demo
    2nd Place for Zhenjiang's poster
    2nd Place for Zhenjiang’s poster
    2nd Place for Zhenjiang's poster
    2nd Place for Zhenjiang’s poster
    2nd Place for Zhenjiang's poster
    2nd Place for Zhenjiang’s poster
    Honorable Mention for Sam's poster
    Honorable Mention for Sam’s poster
    Honorable Mention for Sam's poster
    Honorable Mention for Sam’s poster
    Honorable Mention for Yuang's poster
    Honorable Mention for Yuang’s poster

     

     

     

  • Two surveys: neuro-symbolic AIoT and CPS sustainability
  • Sam, Yuang, Zhenjiang present posters at UF AI Days 2024
    On October 29, 2024, the three students presented posters about the following papers: 
  • Ivan presents calibrated visual safety prediction at TACPS workshop at ESWEEK
    Ivan gave an invited talk “How Safe Will I Be Given What I See? Calibrated Visual Safety Chance Prediction with (Foundation) World Models”. The discussion was very active and generated sufficient questions for the rest of Zhenjiang’s PhD.  Relevant links:  Talk abstract: In safety-critical autonomous systems, safety prediction traditionally relies on low-dimensional data with specific physical meanings, such as poses and velocities. However, such data is not always available, which leaves only high-dimensional sensor observations, such as images from cameras or LiDAR scans, and makes safety prediction increasingly challenging. This talk reports on the recent techniques for using high-dimensional observation data for safety prediction; at the heart of these techniques is the notion of a world model, which can predict future observations without meaningful low-dimensional data. We present several world models implemented with neural representation learning as well as foundation models for image segmentation and natural language prediction. Additionally, we propose a novel uncertainty quantification technique that combines confidence calibration with conformal prediction.
  • Yuang presents high-dimensional reachability at FM 2024

    Yuang Geng presented his work on reachability for vision-based neural-network controllers at the 26th International Symposium on Formal Methods (FM). Reportedly, the attendees are curious about the mapping between states and images. 

    Citation and further materials: 

  • Zhenjiang presents calibrated safety predictors at L4DC 2024
    Zhenjiang Mao presented his work on learning-enabled safety prediction (poster, paper) at the 6th Annual Conference on Learning for Decision and Control (L4DC 2024) in Oxford, UK. Reportedly, the attendees like math more than he does.  Citation: 
  • New NSF project on neuro-symbolic perception in CPS
    The new project is named “Neuro-Symbolic Bridge: From Perception to Estimation & Control“. Its goal is to develop a neuro-symbolic calibration framework to repair the mismatch between perception neural networks and downstream cyber-physical tasks such as state estimation and control. It will be carried out in collaboration with Radoslav Ivanov at RPI.  More information: ECE website, NSF website.  Update: our poster summarizes the first year of progress at the NSF CPS PI meeting. 
  • Ivan spends summer at AFRL as a visiting faculty
    For Summer 2024, Ivan Ruchkin will join the Visiting Faculty Research Program (VFRP) at the Air Force Research Laboratory (AFRL) Information Directorate (RI) in Rome, NY. The program is organized by the Griffis Institute.  Ivan will work on advancing safety verification for high-dimensional controllers. He will also participate in the cyber assurance group’s efforts on testing and assurance for learning-enabled systems. 
  • Zhenjiang presents two papers and a poster at ICRA 2024

    The papers were on foundation world models and language-enhanced OOD. The audience response was, reportedly, positive and encouraged the implementation on physical robotic systems. Citations: 

  • Ivan presents NN repair with preservation at ICCPS 2024
    In the first presentation of ICCPS 2024, Ivan showcased a method to repair a neural network controller while preserving its verification results.  Citation: 
  • Language-enhanced OOD detection: new preprint online

    Our paper gives users of autonomous cars the ability to describe in natural language what conditions they consider nominal or anomalous.

    Citation: 

    • Zhenjiang Mao, Dong-You Jhong, Ao Wang, Ivan Ruchkin. Language-Enhanced Latent Representations for Out-of-Distribution Detection in Autonomous Driving [arxiv], Preprint 2024.
  • F1/10 racing demo for the ECE External Advisory Board
    Industry leaders visited the ECE department to witness the variety of work happening here. Thanks to everyone who helped, especially Carson Sobolewski and Lorant Domokos who led the demonstration.  Some videos and photos from the event:     
  • First batch of students finishes the CURE racing course
    Congratulations to the nine freshmen participants: Ramsey Makan, Jonas Dickens, Tyler Ruble, Christopher Oeltjen, Carter Amaba, Aditya Gandhi, Emilia Delaune, Ethan Krol, and Giancarlo Vidal! And a big thank you to the mentors: Ao Wang, Sam Jhong, Lorant Domokos, and Carson Sobolewski.  More information on this CURE course is here.  
  • Foundation world models: new preprint online

    Our paper develops training-free world models based on foundation models with interpretable latent states.

    Update: presented at the probabilistic robotics workshop at ICRA’24.

    Citation: 

  • TEA Lab moves to Malachowsky Hall
    Now found in Malachowsky 4100, with a brand new racing track coming soon! 
  • Verifying high-dimensional controllers: new preprint online
    Our new draft verifies image-based controllers by approximating them with several low-dimensional ones. Citation: 
    • Yuang Geng, Souradeep Dutta, Ivan Ruchkin. Bridging Dimensions: Confident Reachability for High-Dimensional Controllers [arxiv]. Preprint, in submission.
  • AIAA ASCEND 2023
    Ivan participates in a panel on dependable space autonomy
    Resolving Barriers to Infusion: Fielding Dependable Autonomous Space Systems at AIAA ASCEND 2023.
  • Ivan serves on the PC of ICCPS’24 and AAAI’24
    Consider submitting your papers there.
  • How safe am I given what I see? New preprint online
    Update: a poster was presented at UF AI Days 2023. This paper develops safety chance prediction for image-controlled autonomous systems with calibration guarantees. Citation: 
    • Zhenjiang Mao, Carson Sobolewski, Ivan Ruchkin. How Safe Am I Given What I See? Calibrated Prediction of Safety Chances for Image-Controlled Autonomy [arxiv]. Preprint, in submission. 
  • Invited talk at the DACPS workshop & ETH Autonomy Talks
    Update 1: an extended version of this talk was given at a UF MAE Affiliate Seminar. The recording can be found here (UF login required). Update 2: another version of this walk was given at the ETH Autonomy Talks (video).  Update 3: yet another version of this talks was given as a CNEL Seminar. The talk titled “Verify-then-Monitor: Calibration Guarantees for Safety Confidence” (see the slides here) was presented at the Sixth International Workshop on Design Automation for Cyber-Physical Systems (DACPS), part of the Design Automation Conference (DAC) 2023. Abstract: Autonomous cyber-physical systems (CPS) are increasingly deployed in complex and safety-critical environments. To help CPS interact with such environments, learning-enabled components, such as neural networks, often implement perception and control functions. Unfortunately, the complexity of the environments and learning components is a major challenge to ensuring the safety of CPS. An emerging assurance paradigm prescribes verifying as much of the CPS as possible at design time – and then monitoring the probability of safety at run time in case of unexpected situations. How can we guarantee that the monitor produces a probability that is well-calibrated to the true chance of safety? This talk will overview our recent answers in two settings. The first combines Bayesian filtering with probabilistic model checking of Markov decision processes. The second is based on confidence monitoring of assumptions behind closed-loop neural-network verification.
  • TEA Lab hosts K-12 students for the Robotics-AIoT Visit Day

    On June 15, 2023, the UF ECE Department hosted ~30 school students from the Westwood Middle School and Buchholz High School for a day visit at the Robotics, AI, and IoT research laboratories for educational presentations, research demonstrations, and mentoring discussions. It was a lot of fun for everyone!

    Photos from the event

    Kudos to the other participating labs: SmartDATA Lab, RoboPI Lab, WISE Lab

  • Causal NN controller repair presented at ICAA’23
    Shown above is a 5-step workflow of our causal repair: (1) Extract the behaviors of a learning component as an I/O table. (2) Encode the dependency of the desired property outcome on the I/O behaviors with a Halpern-Pearl model. (3) Search for a counterfactual model value assignment, revealing an actual cause and a repair. (4) Decode the found assignment as a counterfactual component behavior. (5) Replace the original learning component with a repaired component that performs this counterfactual behavior to fix the system. Citation:      
  • Conservative safety monitoring presented at NFM’23
    Shown above is our conservative monitoring approach that leverages probabilistic reachability offline and combines it with calibrated state estimation. Citation:  
  • Red DonkeyCar
    DonkeyCars are racing autonomously
    Our lab is now running neural network-controlled racing cars based on raw camera images: Sometimes things don’t go as planned: Such is the brittle nature of deep learning. We’ll be working on predicting and preventing such accidents.
  • Ivan Ruchkin to serve on the PC of ICCPS’23
    Program Committee of the 14th ACM/IEEE International Conference on Cyber-Physical Systems    
  • TEA Lab is established
    TEA lab’s mission is to develop engineering methodologies for safe autonomous systems that are aware of their own limitations, as illustrated above. More details about this vision can be found in this slide deck.