ICRA 2025
IEEE International Conference on Robotics and Automation | Atlanta | May 19–23 , 2025
Robots are on the cusp of transforming society, powered by bold research and rapid innovation. From embodied AI to adaptive autonomy, Georgia Tech is building machines that learn, collaborate, and have the potential to reshape industries. The future is not science fiction — it’s being engineered today, and it begins here.
Welcome to Georgia Tech at ICRA 2025.
Robots are on the cusp of transforming society, powered by bold research and rapid innovation. From embodied AI to adaptive autonomy, Georgia Tech is building machines that learn, collaborate, and have the potential to reshape industries. The future is not science fiction — it’s being engineered today, and it begins here.
Welcome to Georgia Tech at ICRA 2025.
Smile and say, “ICRA!” Robots showed off their skills at #ICRA2025. So did our researchers presenting at the international event. It’s not just fancy hardware, it’s the future of embodied AI.
Georgia Tech at ICRA 2025
Learn more about robotics at
Georgia Tech’s Institute for Robotics and Intelligent Machines (IRIM) 
By the Numbers
Faculty with Papers
College of Sciences
Georgia Tech Research Institute
Institute for People & Technology
Partner Organizations on Papers
Arizona State University • Boston Dynamics AI Institute • California Institute of Technology • Carnegie Mellon University • Columbia University • Cornell University • Emory University • ETRI • Georgia Tech • Google • Harvard University • Hillsdale College • Honda Research Institute USA • Hong Kong University of Science and Technology • Intuitive Surgical • Jet Propulsion Laboratory • Johns Hopkins University • Massachusetts Institute of Technology • Max Planck Institute for Intelligent Systems • Mercedes-Benz Research & Development North America • Mitsubishi Electric Research Labs • National University of Singapore • New York University • Nuro AI • Nvidia Research • Pusan National University • Rochester Institute of Technology • RWTH Aachen University • Sandia National Labs • Shanghai Jiao Tong University • Sogang University • Stanford University • Technical University of Munich • The Hong Kong University of Science and Technology (Guangzhou) • The Ohio State University • The University of Electro-Communications • Tsinghua University • United States Department of Agriculture – Agricultural Research • Université De Montréal • University of Arkansas • University of California, Berkeley • University of California, Los Angeles • University of Cambridge • University of Delaware • University of Freiburg • University of Illinois Chicago • University of Michigan • University of North Carolina at Charlotte • University of Southern California • University of Texas at Austin • University of the Bundeswehr Munich • University of Toronto • University of Wisconsin, Madison • US Naval Research Laboratory • Wuhan University • Zoox
The Big Picture 
Welcome to ICRA 2025: Advancing the Frontiers of Robotics and Automation
Senior Program Committee, ICRA 2025
School of Interactive Computing, Georgia Tech
Sed ut perspiciatis unde omnis iste natus error sit voluptatem accusantium doloremque laudantium, totam rem aperiam, eaque ipsa quae ab illo inventore veritatis et quasi architecto beatae vitae dicta sunt explicabo.
Nemo enim ipsam voluptatem quia voluptas sit aspernatur aut odit aut fugit, sed quia consequuntur magni dolores eos qui ratione voluptatem sequi nesciunt. Neque porro quisquam est, qui dolorem ipsum quia dolor sit amet, consectetur, adipisci velit, sed quia non numquam eius modi tempora incidunt ut labore et dolore magnam aliquam quaerat voluptatem. Ut enim ad minima veniam, quis nostrum exercitationem ullam corporis suscipit laboriosam, nisi ut aliquid ex ea commodi consequatur? Quis autem vel eum iure reprehenderit qui in ea voluptate velit esse quam nihil molestiae consequatur, vel illum qui dolorem eum fugiat quo voluptas nulla pariatur?
Sed ut perspiciatis unde omnis iste natus error sit voluptatem accusantium doloremque laudantium, totam rem aperiam, eaque ipsa quae ab illo inventore veritatis et quasi architecto beatae vitae dicta sunt explicabo. Nemo enim ipsam voluptatem quia voluptas sit aspernatur aut odit aut fugit, sed quia consequuntur magni dolores eos qui ratione voluptatem sequi nesciunt. Neque porro quisquam est, qui dolorem ipsum quia dolor sit amet, consectetur, adipisci velit, sed quia non numquam eius modi tempora incidunt ut labore et dolore magnam aliquam quaerat voluptatem.
Swarm Highlights
Explore Featured Tech Teams at ICRA 2025
medusai: Interactive AI-driven Robotic Sculpture
ARTS IN ROBOTICS
The AI-driven robotic sculpture medusai is an invited ICRA installation by Gil Weinberg and his team. It responds to and interacts with humans through sound, light, touch, and movement. Inspired by the Greek myth of Medusa, the sculpture features seven robotic arms in the form of “snake hair” installed on top of an 8×10 ft. metallic “face” structure. Human movement around medusai is captured by a top mounted camera, and an artificial vision tracking system drives the robotic arms to follow humans, pluck strings, and hit drums around its surface. medusai also responds with light and electronic sound to humans’ activity such as drumming and plucking strings.
May 19, 7:30 PM – 9:30 PM | The Goat Farm ATL
Advancing Robotics for Complex Aquatic Navigation
ROBOPHYSICAL MODELS
AquaMILR and AquaMILR+ are new untethered limbless robots designed for agile navigation in complex aquatic environments. The robots use a bilateral actuation mechanism, inspired by musculoskeletal actuation in anguilliform swimming organisms, allowing undulatory swimming from head to tail. This actuation is enhanced by mechanical intelligence, improving maneuverability around obstacles.
AquaMILR+ also features a depth control system inspired by swim bladders of eels and sea snakes, offering capabilities that most anguilliform robots lack. Additional features such as fins and a tail improve stability and propulsion efficiency. Tests in open water and indoor aquatic environments highlight their capabilities, positioning it for search and rescue and deep-sea exploration tasks.
TEAM: Tianyu Wang, Matthew Fernandez, Nishanth Mankame, Galen Tunnicliffe, Velin Kojouharov, Donoven Dortilus, Peter Gunnarson, John Dabiri, Daniel Goldman
Next-Gen Robotic Pollination for Indoor Farming
AGRICULTURAL AUTOMATION
Effective pollination is a key challenge for indoor farming, since bees struggle to navigate without the sun. While a variety of robotic system solutions have been proposed, it remains difficult to autonomously check that a flower has been sufficiently pollinated to produce high-quality fruit, which is especially critical for self-pollinating crops such as strawberries.
New agricultural robotics research combines a novel robotic system with an algorithmic approach that can handle delicate flora for indoor farming. The proposed hardware combines a 7-degree-of-freedom (DOF) manipulator arm with a custom end-effector, comprised of an endoscope camera, a 2-DOF microscope subsystem, and a custom vibrating pollination tool; this is paired with algorithms to detect and estimate the pose of strawberry flowers, navigate to each flower, pollinate using the tool, and inspect with the microscope. The key novelty is vibrating the flower from below while simultaneously inspecting with a microscope from above. Each subsystem is validated via extensive experiments.
TEAM: Chuizheng Kong, Alex Qiu, Idris Wibowo, Marvin Ren, Aishik Dhori, Kai-Shu Ling, Ai-Ping Hu, Shreyas Kousik
Improving Multi-Legged Locomotion on Complex Terrains
LEGGED LOCOMOTION
Characterized by their elongate bodies and relatively simple legs, multi-legged robots have the potential to move through complex terrains for applications such as search-and-rescue and terrain inspection. Prior work has developed effective and reliable locomotion strategies for multi-legged robots by propagating the two waves of lateral body undulation and leg stepping, what the research team refers to as the two-wave template. However, these robots have limited capability to climb over obstacles with sizes comparable to their heights, with the team hypothesizing that such limitations stem from the two-wave template.
Seeking effective alternative waves for obstacle-climbing, the researchers designed a five-segment robot with static (non-actuated) legs, where each cable-driven joint has a rotational degree-of-freedom (DoF) in the sagittal plane (vertical wave) and a linear DoF (peristaltic wave).
The research team tested robot locomotion performance on a flat terrain and a rugose terrain. While there were marginal benefits of using peristalsis (wave-like muscle contractions to move the body) on flat-ground locomotion, the inclusion of a peristaltic wave substantially improved the locomotion performance in rugose terrains: it not only enabled obstacle-climbing capabilities with obstacles having a similar height as the robot, but it also significantly improved the traversing capabilities of the robot in such terrains. The results demonstrate an alternative actuation mechanism for multi-legged robots, paving the way towards all-terrain multi-legged robots.
TEAM: Massimiliano Iaschi* (pictured), Baxi Chong*, Tianyu Wang, Jianfeng Lin, Zhaochen Xu, Daniel Soto, Juntao He, Daniel Goldman
*equal contribution
Also at ICRA | Effective Self-Righting Strategies for Elongate Multi-Legged Robots
A Better Understanding of Robot Teleoperator Performance
TELEOPERATION
Advances in robot teleoperation have enabled groundbreaking innovations in many fields, such as space exploration, healthcare, and disaster relief. The human operator’s performance plays a key role in the success of any teleoperation task, with prior evidence suggesting that operator stress and workload can impact task performance. As robot teleoperation is currently deployed in safety-critical domains, it is essential to analyze how different stress and workload levels impact the operator.
New Georgia Tech research represents one of the first studies investigating how both stress and workload impact teleoperation performance. The team conducted a novel study to jointly manipulate users’ stress and workload and analyze the user’s performance through objective and subjective measures. Results indicate that, as stress increased, over 70% of participants performed better up to a moderate level of stress; yet, the majority of participants performed worse as the workload increased. Importantly, the experimental design elucidated that stress and workload have related yet distinct impacts on task performance, with workload mediating the effects of distress on performance.
TEAM: Sam Yi Ting, Erin Hedlund-Botti, Manisha Natarajan, Jamison Heard, Matthew Gombolay (pictured)
Collective Behavior of Entangled Robotic Worms
MULTI-ROBOT SWARMS
Researchers from ETH Zurich, Harvard, and Georgia Tech have combined soft robots and living worms as systems to study collective behaviors driven by physical entanglement. The researchers demonstrated individual and group movement as well as tunable cohesion in both robotic and biological “blobs,” highlighting the generalizability of entanglement-based behaviors. Researchers say the work can be extended to investigate tasks such as collective transport, where entangled groups move objects too large for individuals, and explore mechanisms for controlled disentanglement. By comparing the robot and worm systems, the authors aim to identify key features underlying observed collective behaviors and plan to develop autonomous robot systems with embedded sensing and untethered operation.
The foreground image shows a proof-of-concept demonstration of collective transport with a robot blob. The blob entangles with a 3D-printed bust and carries it along as it undergoes collective
locomotion.
BEST PAPER FINALIST 
TEAM: Carina Kaeser, Junghan Kwon, Elio Challita, Harry Tuazon, Robert Wood, Saad Bhamla, Justin Werfel
FEATURED
New Algorithm Teaches Robots Through Human Perspective
With just 90 minutes of first-person video, a humanoid robot learned household tasks 4x faster — paving the way for scalable, human-taught assistive robots.
FEATURED
New Algorithm Teaches Robots Through Human Perspective
With just 90 minutes of first-person video, a humanoid robot learned household tasks 4x faster — paving the way for scalable, human-taught assistive robots.
By Nathan Dean, School of Interactive Computing
A new data creation paradigm and algorithmic breakthrough from Georgia Tech has laid the groundwork for humanoid assistive robots to help with laundry, dishwashing, and other household chores. The framework enables these robots to learn new skills by mimicking actions from first-person videos of everyday activities.
Current training methods limit robots from being produced at the necessary scale to put a robot in every home, said Simar Kareer, a Ph.D. student in the School of Interactive Computing.
“Traditionally, collecting data for robotics means creating demonstration data,” Kareer said. “You operate the robot’s joints with a controller to move it and achieve the task you want, and you do this hundreds of times while recording sensor data, then train your models. This is slow and difficult. The only way to break that cycle is to detach the data collection from the robot itself.”
Other fields, such as computer vision and natural language processing (NLP), already leverage training data passively culled from the internet to create powerful generative AI and large-language models (LLMs).
Many roboticists, however, have shifted toward interventions that allow individual users to teach their robots how to perform tasks. Kareer believes a similar source of passive data can be established to enable practical generalized training that scales the production of humanoid robots.
This is why Kareer collaborated with School of IC Assistant Professor Danfei Xu and his Robot Learning and Reasoning Lab to develop EgoMimic, an algorithmic framework that leverages data from egocentric videos.
Meta’s Ego4D dataset inspired Kareer’s project. The benchmark dataset, released in 2023, consists of first-person videos of humans performing daily activities. This open-source data set trains AI models from a first-person human perspective.
“When I looked at Ego4D, I saw a dataset that’s the same as all the large robot datasets we’re trying to collect, except it’s with humans,” Kareer said. “You just wear a pair of glasses, and you go do things. It doesn’t need to come from the robot. It should come from something more scalable and passively generated, which is us.”
Kareer acquired a pair of Meta’s Project Aria research glasses, which contain a rich sensor suite and can record video from a first-person perspective through external RGB and SLAM cameras.
Kareer recorded himself folding a shirt while wearing the glasses and repeated the process. He did the same with other tasks such as placing a toy in a bowl and groceries into a bag. Then, he constructed a humanoid robot with pincers for hands and attached the glasses to the top to mimic a first-person viewpoint.
The robot performed each task repeatedly for two hours. Kareer said building a traditional training algorithm would take days of teleoperating and recording robot sensory data. For his project, he only needed to gather a baseline of sensory data to ensure performance improvement.
Kareer bridged the gap between the two training sets with the EgoMimic algorithm. The robot’s task performance rating increased by as much as 400% among various tasks with just 90 minutes of recorded footage. It also showed the ability to perform these tasks in unseen environments.
If enough people wear Aria glasses or other smart glasses while performing daily tasks, it can create the passive data bank needed to train robots on a massive scale.
This type of data collection can enable nearly endless possibilities for roboticists to help humans achieve more in their everyday lives. Humanoid robots can be produced and trained at an industrial level and be able to perform tasks the same way humans do.
“This work is most applicable to jobs that you can get a humanoid robot to do,” Kareer said. “In whatever industry we are allowed to collect egocentric data, we can develop humanoid robots.”
Kareer will present his paper on EgoMimic at the 2025 IEEE Engineers’ International Conference on Robotics and Automation (ICRA), which will take place from May 19 to 23 in Atlanta. The paper was co-authored by Xu and School of IC Assistant Professor Judy Hoffman, fellow Tech students Dhruv Patel, Ryan Punamiya, Pranay Mathur, and Shuo Cheng, and Chen Wang, a Ph.D. student at Stanford.
By Nathan Dean, School of Interactive Computing
A new data creation paradigm and algorithmic breakthrough from Georgia Tech has laid the groundwork for humanoid assistive robots to help with laundry, dishwashing, and other household chores. The framework enables these robots to learn new skills by mimicking actions from first-person videos of everyday activities.
Current training methods limit robots from being produced at the necessary scale to put a robot in every home, said Simar Kareer, a Ph.D. student in the School of Interactive Computing.
“Traditionally, collecting data for robotics means creating demonstration data,” Kareer said. “You operate the robot’s joints with a controller to move it and achieve the task you want, and you do this hundreds of times while recording sensor data, then train your models. This is slow and difficult. The only way to break that cycle is to detach the data collection from the robot itself.”
Other fields, such as computer vision and natural language processing (NLP), already leverage training data passively culled from the internet to create powerful generative AI and large-language models (LLMs).
Many roboticists, however, have shifted toward interventions that allow individual users to teach their robots how to perform tasks. Kareer believes a similar source of passive data can be established to enable practical generalized training that scales the production of humanoid robots.
This is why Kareer collaborated with School of IC Assistant Professor Danfei Xu and his Robot Learning and Reasoning Lab to develop EgoMimic, an algorithmic framework that leverages data from egocentric videos.
Meta’s Ego4D dataset inspired Kareer’s project. The benchmark dataset, released in 2023, consists of first-person videos of humans performing daily activities. This open-source data set trains AI models from a first-person human perspective.
“When I looked at Ego4D, I saw a dataset that’s the same as all the large robot datasets we’re trying to collect, except it’s with humans,” Kareer said. “You just wear a pair of glasses, and you go do things. It doesn’t need to come from the robot. It should come from something more scalable and passively generated, which is us.”
Kareer acquired a pair of Meta’s Project Aria research glasses, which contain a rich sensor suite and can record video from a first-person perspective through external RGB and SLAM cameras.
Kareer recorded himself folding a shirt while wearing the glasses and repeated the process. He did the same with other tasks such as placing a toy in a bowl and groceries into a bag. Then, he constructed a humanoid robot with pincers for hands and attached the glasses to the top to mimic a first-person viewpoint.
The robot performed each task repeatedly for two hours. Kareer said building a traditional training algorithm would take days of teleoperating and recording robot sensory data. For his project, he only needed to gather a baseline of sensory data to ensure performance improvement.
Kareer bridged the gap between the two training sets with the EgoMimic algorithm. The robot’s task performance rating increased by as much as 400% among various tasks with just 90 minutes of recorded footage. It also showed the ability to perform these tasks in unseen environments.
If enough people wear Aria glasses or other smart glasses while performing daily tasks, it can create the passive data bank needed to train robots on a massive scale.
This type of data collection can enable nearly endless possibilities for roboticists to help humans achieve more in their everyday lives. Humanoid robots can be produced and trained at an industrial level and be able to perform tasks the same way humans do.
“This work is most applicable to jobs that you can get a humanoid robot to do,” Kareer said. “In whatever industry we are allowed to collect egocentric data, we can develop humanoid robots.”
Kareer will present his paper on EgoMimic at the 2025 IEEE Engineers’ International Conference on Robotics and Automation (ICRA), which will take place from May 19 to 23 in Atlanta. The paper was co-authored by Xu and School of IC Assistant Professor Judy Hoffman, fellow Tech students Dhruv Patel, Ryan Punamiya, Pranay Mathur, and Shuo Cheng, and Chen Wang, a Ph.D. student at Stanford.
Meet the Team
Simar Kareer
Ph.D. student
Dhruv Patel
Robotics M.S. student
Ryan Punamiya
Computer Science B.S. student
Pranay Mathur
MS Robotics 2024
Shuo Cheng
Ph.D. student
Chen Wang
Ph.D. student, Stanford
Judy Hoffman
Associate Professor, Interactive Computing
Danfei Xu
Associate Professor, Interactive Computing
Bruce Walker Named Founding Director of New Center to Advance Human-AI-Robot Collaboration
Imagine a future where robotic guide dogs lead the visually impaired, flying cars navigate the skies, and electric self-driving vehicles communicate effortlessly with pedestrians.
That future is being shaped today at Georgia Tech’s Center for Human-AI-Robot Teaming (CHART). Led by Bruce Walker, a professor in the School of Psychology and the School of Interactive Computing, the newly launched Center aims to transform how humans, artificial intelligence, and robots work together. By focusing on the dynamic partnership between humans and intelligent systems, CHART will explore how humans can collaborate more effectively with artificial intelligence systems and robots to solve critical scientific and societal challenges.
Walker is a coauthor in the ICRA 2025 technical program on the paper Do Looks Matter? Exploring Functional and Aesthetic Design Preferences for a Robotic Guide Dog.
Robotics World Converges on Atlanta for ICRA 2025
The world’s largest robotics conference is coming to Atlanta, and 136 researchers and students from Georgia Tech will showcase their novel and groundbreaking contributions to a booming field.
The IEEE International Conference on Robotics and Automation (ICRA) will be held Monday through Friday at the Georgia World Congress Center.
“This is the flagship robotics conference,” said Seth Hutchinson, a former Georgia Tech professor who served as one of two general chairs for this year’s event. “Most of the robotics researchers you want to hear from or see will be at this conference.”
ICRA 2025 Expo
Wed, May 21, 3-5 pm
Showcasing Cutting-Edge Robotic Research
Smart Foot System for Enhanced Robot Mobility on Challenging Terrains
Deniz Kerimoglu, Burak Catalbas, Bahadir Catalbas, Daniel Goldman
Robotic platforms mainly focus on locomoting and operating in structured environments such as homes, factory settings, highways and streets etc. These robots can navigate reliably in relatively predictable and controlled settings by relying on predictable ground interactions to perform various tasks. However, robots must also achieve stable locomotion in unpredictable, and challenging terrain such as natural environments and hazardous areas where human operation is difficult, enabling tasks such as exploration, load carrying, and infrastructure maintenance.
Ground Control Robotics ICRA 2025 Demo Proposal
Daniel Soto, Esteban Flores, Daniel Goldman
Multi-legged, undulatory robots possess many advantageous properties for locomotion over unstructured and crowded terrain including low profiles and robustness to missing foot contacts. Despite these advantages, coordinating a high number of legs (6+) and body joints represents a many degree of freedom control problem that has limited their practical and commercial viability. Ground Control Robotics LLC. (GCR), in collaboration with researchers at Georgia Tech, seeks to commercialize these systems for agricultural use by advancing robophysical theories and developing robust mechanical systems.
Multimodal Perception with Legged Mobile Manipulator for Visual, Thermal, and Radiation Monitoring
Hojoon Son, Youndo Do, Marc Zebrowitz, Jacob Faile, Spencer Banks, Myeongjun Choi, Fan Zhang
The proposal presents a multimodal robotic platform for remote visual, thermal, and radiation monitoring in hazardous or unknown environments. The system integrates a Unitree B1 quadruped robot with a Unitree Z1 robotic arm to create a mobile and semi-autonomous perception platform. Equipped with a Teledyne FLIR Hadron 640R, which integrates a long-wave infrared (thermal) camera and a 1080p visible-light imaging sensor, along with an SPRD-ER gamma radiation detector, the robot fuses visual, thermal, and radiation data for real-time monitoring and environmental awareness.
EgoMimic-Expo: Demonstrating Robot Learning from Egocentric Data
Simar Kareer, Dhruv Patel, Ryan Punamiya, Pranay Mathur, Shuo Cheng, Chen Wang, Judy Hoffman, Danfei Xu
This demo showcases EgoMimic, a robotic system that learns from egocentric human data captured by wearable smart glasses. We will demonstrate how robot manipulation skills can be scaled using easily collected human data. This interactive demo features: (1) Live egocentric video streaming from Project Aria glasses capturing human demonstrations.(2) Policy execution on ’Eve’, our low-cost, humanoid-style bimanual robot performing contact-rich tasks (e.g., shirt folding, grocery packing). Project info: https://egomimic.github.io/
Light Following Robophysical Space Rover with Closed-Loop Gait Strategies at Granular Slope
Bahadir Catalbas, Deniz Kerimoglu, Burak Catalbas, Malone Lincoln Hemsley, Daniel Goldman
Exploring extraterrestrial environments requires planetary rovers to gather data and conduct experiments on challenging terrains like steep granular slopes, obstacles, and craters. To overcome these difficulties, modern rovers have leg-like movement systems that can lift, sweep, and spin their wheels; such a mechanism can apply selective substrate fluidization by changing the sweep and spin speed of wheels to generate effective thrust. We utilize a 30 cm-long laboratory-scale robophysical rover model on a tiltable fluidizing testbed containing poppy seeds.
Monday, May 19
9:10 am – 11:30 am
Advanced Manufacturing Pilot Facility
9:15 am – 11:30 am
Georgia Tech Research Institute
9:15 am – 11:30 am
Georgia Tech Hi-bay Robotics and Human-Augmentation Space
11:40 am – 2:00 pm
Klaus Building Robotics Lab Tour
12:45 pm – 3:00 pm
Advanced Manufacturing Pilot Facility 2
12:45 pm – 3:00 pm
Georgia Tech Research Institute 2
12:45 pm – 3:00 pm
Georgia Tech Hi-bay Robotics and Human-Augmentation Space 2
1:00 pm – 3:30 pm
Klaus Building Robotics Lab Tour 2
RESEARCH
ICRA 2025 Papers with Georgia Tech coauthors are listed and sorted alphabetically by session.
Search complete program details by day and keyword, e.g. “Georgia Tech,” for other work, including Late-Breaking Results,
or see Author Index. Also check out the Workshop listing.
Aerial Robots: Mechanics and Control
Dense Fixed-Wing Swarming Using Receding-Horizon NMPC
Varun Madabushi, Yocheved Kopel, Adam Polevoy, Joseph Moore
This paper presents a method for controlling agile fixed-wing aerial vehicles flying closely together in a swarm. The method uses receding-horizon nonlinear model predictive control (NMPC) to plan dynamic maneuvers while avoiding inter-agent collisions.
Agricultural Automation
Robotic 3D Flower Pose Estimation for Small-Scale Urban Farms
Venkata Harsh Suhith Muriki, Hong Ray Teo, Ved Sengupta, Ai-Ping Hu
This paper presents a novel approach for flower pose estimation using a FarmBot platform with a custom camera end-effector to automate plant phenotyping. By leveraging 3D point cloud data, the system generates 2D images corresponding to six orthogonal viewpoints of the flower.
Agricultural Automation
Towards Closing the Loop in Robotic Pollination for Indoor Farming Via Autonomous Microscopic Inspection
Chuizheng Kong, Alex Qiu, Idris Wibowo, Marvin Ren, Aishik Dhori, Kai-Shu Ling, Ai-Ping Hu, Shreyas Kousik
This work proposes a novel robotic system for indoor farming. The proposed hardware combines a 7-degree-of-freedom (DOF) manipulator arm with a custom end-effector, comprised of an endoscope camera, a 2-DOF microscope subsystem, and a custom vibrating pollination tool.
Assistive Robotics
Do Looks Matter? Exploring Functional and Aesthetic Design Preferences for a Robotic Guide Dog
Aviv Cohav, Xinran Gong, Joanne Taery Kim, Clint Zeagler, Sehoon Ha, Bruce Walker
Dog guides offer an effective mobility solution for blind or visually impaired (BVI) individuals, but conventional dog guides have limitations including the need for care, potential distractions, societal prejudice, high costs, and limited availability.
Bioinspiration and Biomimetics
AquaMILR: Mechanical Intelligence Simplifies Control of Undulatory Robots in Cluttered Fluid Environments
Tianyu Wang, Nishanth Mankame, Matthew Fernandez, Velin Kojouharov, Daniel Goldman
This paper explores how mechanical intelligence—the idea that physical body mechanics can simplify control—applies to undulatory robots swimming in cluttered aquatic environments.
Bioinspiration and Biomimetics
AquaMILR+: Design of an Untethered Limbless Robot for Complex Aquatic Terrain Navigation
Matthew Fernandez, Tianyu Wang, Galen Tunnicliffe, Donoven Dortilus, Peter Gunnarson, John Dabiri, Daniel Goldman
This paper presents AquaMILR+, an untethered limbless robot designed for agile navigation in complex aquatic environments. The robot uses a bilateral actuation mechanism, inspired by musculoskeletal actuation in anguilliform swimming organisms, allowing undulatory swimming from head to tail.
Bioinspiration and Biomimetics
Bird-Inspired Tendon Coupling Improves Paddling Efficiency by Shortening Phase Transition Times
Jianfeng Lin, Zhao Guo, Alexander Badri-Spröwitz
This paper explores the design of drag-based swimming vehicles, inspired by the coupling tendons of aquatic birds. The challenge of transitioning between the recovery and power phases in swimming is addressed by incorporating tendon coupling mechanisms.
Bio-Inspired Robot Learning
Materials Matter: Investigating Functional Advantages of Bio-Inspired Materials Via Simulated Robotic Hopping
Andrew Schulz, Ayah Ahmad, Maegan Tucker
This paper explores how material properties impact robot performance, inspired by how natural systems use varied materials to their advantage. Using a simulated single-limb hopping robot, the authors tested different material profiles.
Design and Control
Continuously Variable Transmission and Stiffness Actuator Based on Actively Variable Four-Bar Linkage for Highly Dynamic Robot Systems
Jungwoo Hur, Hangyeol Song, Seokhwan Jeong
This paper presents a novel actuation mechanism that combines a continuously variable transmission (CVT) mechanism with a variable stiffness actuator (VSA) for highly dynamic robot systems such as legged robots.
Diffusion for Manipulation
Legibility Diffuser: Offline Imitation for Intent Expressive Motion
Matthew Bronars, Shuo Cheng, Danfei Xu
This paper introduces Legibility Diffuser, a diffusion-based policy for generating intent-expressive motion in human-robot collaboration. Unlike classical motion planners, this approach learns directly from offline human demonstration data.
Diffusion Models
Learning Diverse Robot Striking Motions with Diffusion Models and Kinematically Constrained Gradient Guidance
Kin Man Lee, Sean Ye, Qingyu Xiao, Zixuan Wu, Zulfiqar Zaidi, David D’Ambrosio, Pannag Sanketi, Matthew Gombolay
This paper presents a novel approach to robot learning, focusing on learning diverse robot striking motions for tasks such as table tennis. The authors propose a diffusion modeling approach that is offline and constraint-guided.
Explainable AI in Robotics
CE-MRS: Contrastive Explanations for Multi-Robot Systems
Ethan Schneider, Daniel Wu, Devleena Das, Sonia Chernova
As multi-robot systems grow in complexity, their decisions often become hard for humans to understand. This paper introduces CE-MRS, a framework for generating contrastive explanations—natural language explanations that answer questions like “Why did the system do this instead of that?”
ID and Estimation for Legged Robots
Simultaneous Collision Detection and Force Estimation for Dynamic Quadrupedal Locomotion
Ziyi Zhou, Stefano Di Cairano, Yebin Wang, Karl Berntorp
In this paper we address the simultaneous collision detection and force estimation problem for quadrupedal locomotion using joint encoder information and the robot dynamics only. We design an interacting multiple-model Kalman filter (IMM-KF) that estimates the external force exerted on the robot and multiple possible contact modes.
Imitation Learning
Learning Wheelchair Tennis Navigation from Broadcast Videos with Domain Knowledge Transfer and Diffusion Motion Planning
Zixuan Wu, Zulfiqar Zaidi, Adithya Patil, Qingyu Xiao, Matthew Gombolay
In this paper, we propose a novel and generalizable zero-shot knowledge transfer framework that distills expert sports navigation strategies from web videos into robotic systems with adversarial constraints and out-of-distribution image trajectories.
Imitation Learning for Manipulation
EgoMimic: Scaling Imitation Learning Via Egocentric Video
Simar Kareer, Dhruv Patel, Ryan Punamiya, Pranay Mathur, Shuo Cheng, Chen Wang, Judy Hoffman, Danfei Xu
A major challenge in imitation learning is the need for large, diverse demonstration data. EgoMimic is a full-stack framework designed to scale robotic manipulation using egocentric human videos.
Imitation Learning for Manipulation
Learning Prehensile Dexterity by Imitating and Emulating State-Only Observations
Yunhai Han, Zhenyang Chen, Kyle Williams, Harish Ravichandar
Humans often learn physical skills by first observing experts and then emulating them through practice. Inspired by this, the authors propose CIMER (Combining IMitation and Emulation for Motion Refinement) — a two-stage learning framework for dexterous prehensile manipulation from state-only observations.
Imitation Learning for Manipulation
RoCoDA: Counterfactual Data Augmentation for Data-Efficient Robot Learning from Demonstrations
Ezra Ameperosa, Jeremy Collins, Mrinal Jain, Animesh Garg
RoCoDA is a data augmentation framework for imitation learning that improves generalization and efficiency by combining three key ideas: invariance, equivariance, and causality. It makes synthetic demonstrations by (1) modifying task-irrelevant parts of the scene (causal invariance), and (2) applying rigid SE(3) transformations to objects and adjusting the actions (equivariance).
In-Hand Manipulation
Diffusion-Informed Probabilistic Contact Search for Multi-Finger Manipulation
Abhinav Kumar, Thomas Power, Fan Yang, Sergio Aguilera, Soshi Iba, Rana Soltani Zarrin, Dmitry Berenson
Contact-rich manipulation with multi-fingered hands is difficult due to the complex and hybrid nature of dynamics. This paper introduces DIPS (Diffusion-Informed Probabilistic Search), a planning method that uses A* search informed by a diffusion model trained on high-quality demonstrations.
Integrating Motion Planning/Learning
CLIMB: Language-Guided Continual Learning for Task Planning with Iterative Model Building
Walker Byrnes, Miroslav Bogdanovic, Avi Balakirsky, Stephen Balakirsky, Animesh Garg
Intelligent and reliable task planning is a core capability for generalized robotics, which requires a descriptive domain representation that sufficiently models all object and state information for the scene. We present CLIMB, a continual learning framework for robot task planning.
Learning for Manipulation
Catch It! Learning to Catch in Flight with Mobile Dexterous Hands
Yuanhang Zhang, Tianhai Liang, Zhenyang Chen, Yanjie Ze, Huazhe Xu
This paper tackles the challenge of catching objects thrown through the air—a skill that demands precise, agile, and full-body control. The team built a mobile robot system with a mobile base, a 6-DoF arm, and a 12-DoF dexterous hand.
Legged Locomotion: Novel Platforms
Addition of a Peristaltic Wave Improves Multi-Legged Locomotion Performance on Complex Terrains
Massimiliano Iaschi, Baxi Chong, Tianyu Wang, Jianfeng Lin, Zhaochen Xu, Daniel Soto, Juntao He, Daniel Goldman
Characterized by their elongate bodies and relatively simple legs, multi-legged robots have the potential to locomote through complex terrains for applications such as search-and-rescue and terrain inspection. Prior work has developed effective and reliable locomotion strategies for multi-legged robots by propagating the two waves of lateral body undulation and leg stepping, which we will refer to as the two-wave template.
Legged Locomotion: Novel Platforms
Berkeley Humanoid: A Research Platform for Learning-Based Control
Qiayuan Liao, Bike Zhang, Xuanyu Huang, Xiaoyu Huang, Zhongyu Li, Koushil Sreenath
We introduce Berkeley Humanoid, a reliable and low-cost mid-scale humanoid research platform for learning-based control. Our lightweight, in-house-built robot is designed specifically for learning algorithms with accurate simulation, low simulation complexity, anthropomorphic motion, and high reliability against falls.
Legged Locomotion: Novel Platforms
Effective Self-Righting Strategies for Elongate Multi-Legged Robots
Erik Teder, Baxi Chong, Juntao He, Tianyu Wang, Massimiliano Iaschi, Daniel Soto, Daniel Goldman
Centipede-like robots offer an effective and robust solution to navigation over complex terrain with minimal sensing. However, when climbing over obstacles, such multi-legged robots often elevate their center-of-mass into unstable configurations, where even moderate terrain uncertainty can cause tipping over.
Manipulation Planning and Control
Is Linear Feedback on Smoothed Dynamics Sufficient for Stabilizing Contact-Rich Plans?
Yuki Shirai, Tong Zhao, Hyung Ju Terry Suh, Huaijiang Zhu, Xinpei Ni, Jiuguang Wang, Max Simchowitz, Tao Pang
This paper explores the challenges of designing planners and controllers for contact-rich manipulation, where contact disrupts the smoothness assumptions of many gradient-based controller synthesis tools. The authors analyze the effectiveness of linear feedback control for smoothed dynamics, using contact smoothing to approximate non-smooth systems.
Marine Robotics
A Data-Driven Velocity Estimator for Autonomous Underwater Vehicles Experiencing Unmeasurable Flow and Wave Disturbance
Jinzhi Cai, Scott Mayberry, Huan Yin, Fumin Zhang
Autonomous Underwater Vehicles (AUVs) encounter significant challenges in confined spaces like ports and testing tanks, where vehicle-environment interactions, such as wave reflections and unsteady flows, introduce complex, time-varying disturbances. Model-based state estimation methods can struggle to handle these dynamics, leading to localization errors.
Mechanism Design and Control
Guaranteed Reach-Avoid for Black-Box Systems through Narrow Gaps Via Neural Network Reachability
Long Kiu Chung, Wonsuhk Jung, Srivatsank Pullabhotla, Parth Kishor Shinde, Yadu Krishna Sunil, Saihari Kota, Luis F. W. Batista, Cedric Pradalier, Shreyas Kousik
In the classical reach-avoid problem, autonomous mobile robots are tasked to reach a goal while avoiding obstacles. However, it is difficult to provide guarantees on the robot’s performance when the obstacles form a narrow gap and the robot is a black-box (i.e. the dynamics are not known analytically, but interacting with the system is cheap).
Mechanism Design and Control
RAIL: Reachability-Aided Imitation Learning for Safe Policy Execution
Wonsuhk Jung, Dennis Anthony, Utkarsh Mishra, Nadun Ranawaka Arachchige, Matthew Bronars, Danfei Xu, Shreyas Kousik
Imitation learning (IL) has shown great success in learning complex robot manipulation tasks. However, there remains a need for practical safety methods to justify widespread deployment.
Medical Robot Systems
Design and Modeling of a Compact Spooling Mechanism for the COAST Guidewire Robot
Timothy A. Brumfiel, Jared Grinberg, Betina Siopongco, Jaydev P. Desai
The treatment of many intravascular procedures begins with a clinician manually placing a guidewire to the target lesion to aid in placing other devices. Manually steering the guidewire is challenging due to the lack of direct tip control and the high tortuosity of vessel structures, potentially resulting in vessel perforation or guidewire fracture.
Model Control, Legged Robots
Terrain-Aware Model Predictive Control of Heterogeneous Bipedal and Aerial Robot Coordination for Search and Rescue Tasks
Abdulaziz Shamsah, Jesse Jiang, Ziwon Yoon, Samuel Coogan, Ye Zhao
This study presents a task and motion planning framework for search and rescue operations using a heterogeneous robot team composed of humanoids and aerial robots. A terrain-aware Model Predictive Controller (MPC) is proposed, incorporating terrain elevation gradients learned using Gaussian processes (GP).
Motion Planning
Propagative Distance Optimization for Motion Planning
Yu Chen, Jinyun Xu, Yilin Cai, Ting-Wei Wong, Zhongqiang Ren, Howie Choset, Guanya Shi
This paper focuses on the motion planning problem for serial articulated robots with revolute joints under kinematic constraints. Many motion planners leverage iterative local optimization methods but are often trapped in local minima due to non-convexity of the problem.
Multi-Robot Exploration
Communication-Aware Iterative Map Compression for Online Path-Planning
Evangelos Psomiadis, Ali Reza Pedram, Dipankar Maity, Panagiotis Tsiotras
This paper addresses the problem of optimizing communicated information among heterogeneous, resource-aware robot teams to facilitate their navigation. A mobile robot compresses its local map to assist another robot in reaching a target within an uncharted environment.
Multi-Robot Swarms
Best Paper Finalist
Individual and Collective Behaviors in Soft Robot Worms Inspired by Living Worm Blobs
Carina Kaeser, Junghan Kwon, Elio Challita, Harry Tuazon, Robert Wood, Saad Bhamla, Justin Werfel
California blackworms constitute a recently identified animal system exhibiting unusual collective behaviors, in which dozens to thousands of worms entangle to form a “blob” capable of actions like locomotion as an aggregate. In this paper we describe a system of pneumatic soft robots inspired by the blackworms.
Multi-Robot Systems
A Streamlined Heuristic for the Problem of Min-Time Coverage in Constricted Environments (I)
Young-In Kim, Spiridon Reveliotis
This paper tackles the minimum-time coverage problem for robotic fleets operating in constricted, structured environments, like pipes or narrow service tunnels. It introduces a streamlined heuristic that reduces computation time and a local search method that refines the solution for better performance and scalability.
Multi-Robot Systems
Integrating Multi-Robot Adaptive Sampling and Informative Path Planning for Spatiotemporal Natural Environment Prediction
Siva Kailas, Srujan Deolasee, Wenhao Luo, Woojun Kim, Katia Sycara
This work presents a decentralized framework for multi-robot adaptive sampling and informative path planning to predict spatiotemporal environmental processes. Using Gaussian Process models and peer-to-peer coordination, the system identifies informative sampling locations and plans efficient paths under constraints.
Multi-Robot Systems
Residual Descent Differential Dynamic Game (RD3G) – a Fast Newton Solver for Constrained General Sum Games
Zhiyuan Zhang, Panagiotis Tsiotras
This paper introduces RD3G, a Newton-based solver for multi-agent differential dynamic games with constraints. It dynamically handles constraints and uses efficient techniques to achieve 4× faster performance and 2× higher convergence compared to previous methods.
Novel Methods for Mapping/Localization
Evaluating Global Geo-Alignment for Precision Learned Autonomous Vehicle Localization Using Aerial Data
Yi Yang, Xuran Zhao, Haicheng Charles Zhao, Shumin Yuan, Samuel Bateman, Tiffany A. Huang, Chris Beall, Will Maddern
This paper investigates aligning aerial data with sensor data for autonomous vehicle localization. Evaluated on a 1600 km dataset, the method achieves sub-0.3 meter and 0.5° errors, showing promise for precise, scalable localization.
Offroad Navigation
Dynamics Modeling Using Visual Terrain Features for High-Speed Autonomous Off-Road Driving
Jason Gibson, Anoushka Alavilli, Erica Tevere, Evangelos Theodorou, Patrick Spieler
This paper presents a hybrid dynamics model that uses visual terrain features—extracted via a foundation model like DINOv2—to improve high-speed autonomous off-road driving. The method enables real-time planning and is validated on diverse terrain data from the DARPA RACER program.
Optimization and Optimal Control
Second-Order Stein Variational Dynamic Optimization
Yuichiro Aoyama, Peter Lehmann, Evangelos Theodorou
The authors introduce a new optimization algorithm—Stein Variational Differential Dynamic Programming—that merges sampling-based and gradient-based approaches to improve trajectory optimization. It performs well in Model Predictive Control tasks, offering better convergence and avoiding local minima.
Perception for Mobile Robots
DreamDrive: Generative 4D Scene Modeling from Street View Images
Jiageng Mao, Boyi Li, Boris Ivanovic, Yuxiao Chen, Yan Wang, Yurong You, Chaowei Xiao, Danfei Xu, Marco Pavone, Yue Wang
DreamDrive synthesizes 3D-consistent, 4D driving scenes from street view images using generative video diffusion models and hybrid Gaussian representations. It produces realistic and generalizable driving videos from in-the-wild data, enhancing autonomous perception and planning.
Reinforcement Learning
Learning a High-Quality Robotic Wiping Policy Using Systematic Reward Analysis and Visual-Language Model Based Curriculum
Yihong Liu, Dongyeop Kang, Sehoon Ha
This work improves robotic wiping through a bounded reward formulation and a VLM-based curriculum that guides learning. The combined approach trains high-quality wiping policies across varied surfaces, solving convergence issues faced by standard Deep RL methods.
Reinforcement Learning
PrivilegedDreamer: Explicit Imagination of Privileged Information for Rapid Adaptation of Learned Policies
Morgan Byrd, Jackson Crandell, Mili Das, Jessica Inman, Robert Wright, Sehoon Ha
PrivilegedDreamer is a model-based reinforcement learning framework for hidden-parameter MDPs. It uses a dual recurrent architecture to estimate hidden environment parameters and conditions its networks on them, significantly improving sim-to-real transfer and outperforming state-of-the-art methods on multiple tasks.
Reinforcement Learning Applications
Learning Multi-Agent Coordination for Replenishment at Sea
Byeolyi Han, Minwoo Cho, Letian Chen, Rohan Paleja, Zixuan Wu, Sean Ye, Esmaeil Seraj, David Sidoti, Matthew Gombolay
This work introduces Marine, a MARL environment simulating sea-based logistics with real wave data. Their SchedHGNN model combines a heterogeneous graph neural network and intrinsic rewards to improve coordination under dynamic weather, achieving up to 37.8% better performance than prior baselines.
Representation Learning
Learning Dynamics of a Ball with Differentiable Factor Graph and Roto-Translational Invariant Representations
Qingyu Xiao, Zixuan Wu, Matthew Gombolay
This paper proposes a differentiable factor graph and roto-translational invariant representations to model fast, nonlinear ball dynamics in games like ping pong. The approach achieves improved accuracy with low RMSE and supports agile robot planning in dynamic, contact-rich environments.
Representation Learning
MI-HGNN: Morphology-Informed Heterogeneous Graph Neural Network for Legged Robot Contact Perception
Daniel Chase Butterfield, Sandilya Sai Garimella, NaiJen Cheng, Lu Gan
MI-HGNN is a graph neural network informed by robot morphology for legged robot contact perception. It outperforms a leading baseline by 8.4% using only 0.21% of its parameters, and generalizes to other multi-body systems. Code is available on GitHub: Morphology-Informed-HGNN.
Resiliency and Security
Affine Transformation-Based Perfectly Undetectable False Data Injection Attacks on Remote Manipulator Kinematic Control with Attack Detector
Jun Ueda, Jacob Blevins
This paper demonstrates the viability of perfectly undetectable affine transformation attacks against robotic manipulators. The attacker can implement these communication line attacks by satisfying three conditions presented in this work.
Resiliency and Security
Perfectly Undetectable False Data Injection Attacks on Encrypted Bilateral Teleoperation System Based on Dynamic Symmetry and Malleability
Hyukbin Kwon, Hiroaki Kawase, Heriberto Andres Nieves-Vazquez, Kiminao Kogiso, Jun Ueda
This paper investigates the vulnerability of bilateral teleoperation systems to perfectly undetectable False Data Injection Attacks (FDIAs). The paper focuses on a specific class of cyberattacks: perfectly undetectable FDIAs, where attackers alter signals without leaving detectable traces at all.
Robot Safety
Dynamic Gap: Safe Gap-Based Navigation in Dynamic Environments
Maxwell Asselmeier, Dhruv Ahuja, Abdel Zaro, Ahmad Abuaish, Ye Zhao, Patricio Vela
This paper extends the family of gap-based local planners to unknown dynamic environments by generating provably collision-free properties for hierarchical navigation systems. Unlike existing planners that rely on empirical robustness for dynamic obstacle avoidance, this method performs a formal analysis of dynamic obstacles.
SLAM
HDPlanner: Advancing Autonomous Deployments in Unknown Environments through Hierarchical Decision Networks
Jingsong Liang, Yuhong Cao, Yixiao Ma, Hanqi Zhao, Guillaume Adrien Sartoretti
In this paper, we introduce HDPlanner, a deep reinforcement learning (DRL) based framework designed to tackle two core and challenging tasks for mobile robots: autonomous exploration and navigation. Specifically, HDPlanner relies on novel hierarchical attention networks to empower the robot to reason about its belief across multiple spatial scales.
Soft Robotic Grasping
Kinetostatics and Retention Force Analysis of Soft Robot Grippers with External Tendon Routing
Anthony Gunderman, Yifan Wang, Benjamin Gunderman, Alex Qiu, Milad Azizkhani, Joseph Sommer, Yue Chen
Soft robots (SR) are a class of continuum robots that enable safe human interaction with task versatility beyond rigid robots. This letter presents a kinetostatic modeling approach based on strain energy minimization subject to mechanics and geometric constraints for shape estimation of SR grippers with external tendon routing (ETR).
Software Tools
A Survey on Small-Scale Testbeds for Connected and Automated Vehicles and Robot Swarms
Armin Mokhtarian, Jianye Xu, Patrick Scheffe, Maximilian Kloock, Simon Schäfer, Heeseung Bang, Viet-Anh Le, Sangeet Ulhas, Johannes Betz, Sean Wilson, Spring Berman, Liam Paull, Amanda Prorok, Bassam Alrifaee
This work serves to facilitate researchers’ efforts in identifying existing small-scale testbeds suitable for their experiments and provide insights for those who want to build their own for connected and automated vehicles and robot swarms.
Surgical Robotics: Catheters/Needles
Model-Based Parameter Selection for a Steerable Continuum Robot — Applications to Bronchoalveolar Lavage (BAL)
Amber K. Rothe, Timothy A. Brumfiel, Revanth Konda, Kirsten Williams, Jaydev P. Desai
Bronchoalveolar lavage (BAL) is a minimally invasive procedure for diagnosing lung infections and diseases. Continuum robots could improve the navigation of catheters, guidewires, and endoscopes in such procedures.
Surgical Robotics: Planning
SuFIA-BC: Generating High Quality Demonstration Data for Visuomotor Policy Learning in Surgical Subtasks
Masoud Moghani, Nigel Nelson, Mohamed Ghanem, Andres Diaz-Pinto, Kush Hari, Mahdi Azizian, Ken Goldberg, Sean Huver, Animesh Garg
Behavior cloning facilitates the learning of dexterous manipulation skills, yet the complexity of surgical environments, the difficulty and expense of obtaining patient data, and robot calibration errors present unique challenges for surgical robot learning. We present SuFIA-BC: visual Behavior Cloning policies for Surgical First Interactive Autonomy Assistants.
Task and Motion Planning
Optimization-Based Task and Motion Planning under Signal Temporal Logic Specifications Using Logic Network Flow
Xuan Lin, Jiming Ren, Samuel Coogan, Ye Zhao
This paper proposes an optimization-based task and motion planning framework, called Logic Network Flow, to integrate signal temporal logic (STL) specifications into efficient mixed-binary linear programs. Logic Network Flow encodes temporal predicates as polyhedral constraints on edges of a network flow.
Teleoperation
The Impact of Stress and Workload on Human Performance in Robot Teleoperation Tasks
Sam Yi Ting, Erin Hedlund-Botti, Manisha Natarajan, Jamison Heard, Matthew Gombolay
Advances in robot teleoperation have enabled groundbreaking innovations in many fields, such as space exploration, healthcare, and disaster relief. The human operator’s performance plays a key role in the success of any teleoperation task.
Testing and Validation
Learning-Based Bayesian Inference for Testing of Autonomous Systems
Anjali Parashar, Ji Yin, Charles Dawson, Panagiotis Tsiotras, Chuchu Fan
For the safe operation of robotic systems, it is important to accurately understand its failure modes using prior testing. Hardware testing of robotic infrastructure is known to be slow and costly.
Vision-Based Navigation
Safer Gap: Safe Navigation of Planar Nonholonomic Robots with a Gap-Based Local Planner
Shiyu Feng, Ahmad Abuaish, Patricio Vela
This paper extends the idea of gap-based robot navigation to nonholonomic robots with safety guarantees. The authors propose a safe navigation technique that ensures robot movement in dynamic environments while guaranteeing collision avoidance.
Vision-Based Navigation
X-MOBILITY: End-To-End Generalizable Navigation Via World Modeling
Wei Liu, Huihua Zhao, Chenran Li, Joydeep Biswas, Billy Okal, Pulkit Goyal, Yan Chang, Soha Pouya
This paper introduces xmobility, an end-to-end generalizable navigation model that overcomes existing challenges by leveraging three key ideas. First, xmobility employs an auto-regressive world modeling architecture with a latent state space to capture world dynamics.
