University of Michigan Robotics Department

A summary of what's been happening the last few months around Michigan Robotics in our latest substack newsletter, including: • Enhancing actuator design for humanoid robots • How restaurant employees view working alongside robots • A one-of-a-kind collaborative teaching model Read (and subscribe to not miss the next one): https://lnkd.in/gajhYwNv

What are faculty looking for in a PhD application? The Robotics PhD application deadline is December 1 (and master's on January 15). We took the timely opportunity to ask faculty what they look for in an application. Patricia Alves-Oliveira: I’m looking for a candidate who can communicate who they are, what they have achieved, and where they see themselves in the future. Even if they do not align with what the applicant writes, this reveals how they think and can drive my interest. I look for candidates who go beyond buzzwords and research hypes and instead reveal insight. The candidate should be able to describe projects they led or were a central part of, detail how their role in the project contributed to its fruition, and describe challenges they faced and how they overcame them (or explain why they didn’t). My lab is focused on human-robot interaction – the word “interaction” is key and I try to understand who the applicant is and how they can complement my lab at a research and human levels. Steven Ceron: I am looking for students that demonstrate initiative, creativity, and hard work. A specific skill set level is not necessary, but I do become more interested in students that have used their time to pursue difficult projects (robotics-related would be a plus) and carry them out to the intended goal; this could be independently, through research, design teams, etc. When reading the personal statements, I would like to see that a student can communicate their ideas clearly and that through the description of their past projects there comes a very noticeable interest in improving their skillsets. I am also excited about unusual students that have interesting stories to tell about where they came from and how they became interested in the projects they worked on; the type of students that I can tell want to grow through the projects they work on, are interested in helping others, and with whom I could learn many new things and expand my own research goals. Just like everyone else, I want to see that students are authentic about their background, their motivations, and their goals; the authenticity will show one way or another and it will help students find the right lab. I will also take a close look at recommendation letters from individuals that have worked with the student and can speak about their ability to work towards engineering solutions on their own as well as within a team. For more faculty's answers, see: https://lnkd.in/gu2avxB3

Fantastic talk by Kira Barton from the Mechanical Engineering Department - University of Michigan and the University of Michigan Robotics Department! Learning about her research on Iterative Learning Control (ILC) applied to human-autonomy interactions was a great pleasure! Photos: Control Seminar Series hosted by the Center for Control and Dynamics Systems (CCDS) at UC San Diego Mechanical and Aerospace Engineering Department. Check our website: https://lnkd.in/gy2Zw37q

Hardik Parwana joins the distinguished group of those who defended their University of Michigan Robotics PhD. You can watch the whole defense, "Online Adaptation for Safe Control of Constrained Dynamical Systems": https://lnkd.in/gupFm5zJ In it, Parwana discusses developments to help robots navigate untested environments while maintaining safety, including: • Tools for adapting parametric controllers to satisfy multiple state constraints and improve performance. • A numerical scheme for predicting future states in systems with stochastic dynamics for use in model-predictive control. • A method to assist in decision-making when it is not feasible to satisfy all constraints simultaneously. • Focus on safety-critical control method: control barrier functions (CBF) and introduction of Rate-Tunable CBFs for time-varying parameters and online tuning.

❗❗ Last days to apply for a Postdoc in Human-Robot Interaction (HRI) & Arts University of Michigan Robotics Department. Applications need to be submitted through the following website: https://lnkd.in/d546DEfW **Deadline: 30 Nov 2024**

I’m excited to share that we’ve begun operating our #autonomousvehicles-- with no safety driver-- in Ann Arbor. This marks our second city in under a year, following our Sun City, Arizona deployment. Like Sun City, our Ann Arbor route includes a wide range of challenging driving conditions: unprotected turns, traffic lights, stop signs, pedestrians, and more. But Ann Arbor adds (much!) more weather, greater traffic density, narrower lanes, and higher operating speeds. Over the next few months, we will explore expanding our driver-out coverage area within the city and make plans to transition some of our autonomous vehicle services to #driver-out operations over the coming years.

AI stakeholders are anticipating changes with Donald Trump’s return to the White House. I recently spoke with Newsweek's Marie Boran  to discuss the potential impact on innovation, regulation, and trust in artificial intelligence. https://lnkd.in/dA6_A5vi Some highlights: "In some sense, the status quo for AI regulation is that there is none." Regulations like the Biden administration's AI executive order were starting a critical conversation around safe and trustworthy AI — one that must continue. Innovation in AI is thriving, and I believe it will persist regardless of federal regulations. But "what we need to see is more companies, more leaders stepping up to share the work openly." Transparency fosters not only trust but also accelerates innovation. My vision remains clear that "an open ecosystem for AI, means safer, faster innovation." Sharing models, data, and insights openly creates immense value for the field and society. As the future of AI policy takes shape, I hope to see a continued focus on cultivating trust and collaboration. Let’s work together to build an AI ecosystem that drives both progress and responsibility. #ArtificialIntelligence #Innovation #AIRegulation

As part of IDEAL Institute's Workshop on Navigating the Ethics of Human-AI Collaboration, Lionel Robert joins a panel on AI Agents in Business. Wednesday, 11/20 at 10:30am. More on the workshop: The rapid integration of AI systems into various domains of human life has sparked both enthusiasm and concern. While some envision unprecedented productivity gains, economic growth, and improved healthcare access, others worry about unfair discrimination, workforce displacement, and threats to human autonomy. Treating AI as a tool for augmenting human capabilities rather than replacing them, could reap the benefits of AI while preserving human agency and values. AI Agents (i.e., AI systems that can pursue complex goals with limited direct supervision) hold great promises for the augmentation of human capabilities. However, the emergence of more sophisticated and capable AI agents, which may play increasingly large roles in human lives, raises new questions about the nature of this augmentation and the evolving relationship between humans and AI. And register: https://lnkd.in/gfzG-hQi

Earlier this year, Nima Fazeli recieved a National Science Foundation (NSF) CAREER Award. We took the chance to catch up with Professor Fazeli to learn a bit more about the fascinating research on "intelligent and dexterous robots that seamlessly integrate vision and touch.”

LUCCa is a new approach that tackles an important robotics challenge: accurately quantifying the uncertainty in a robot's dynamics with provable guarantees. The paper, “Quantifying Aleatoric and Epistemic Dynamics Uncertainty via Local Conformal Calibration,” from Luís Marques and Dmitry Berenson in the ARM Lab, describes LUCCa. It considers both aleatoric and epistemic uncertainty, providing a comprehensive understanding of the robot's behavior in environments with perturbations and changes in actual system dynamics. Aleatoric uncertainty: This refers to the inherent randomness in the system, which is already addressed by many existing methods. Epistemic uncertainty: This refers to uncertainty arising from a lack of information or data. Abstract: Whether learned, simulated, or analytical, approximations of a robot’s dynamics can be inaccurate when encountering novel environments. Many approaches have been proposed to quantify the aleatoric uncertainty of such methods, i.e. uncertainty resulting from stochasticity, however these estimates alone are not enough to properly estimate the uncertainty of a model in a novel environment, where the actual dynamics can change. Such changes can induce epistemic uncertainty, i.e. uncertainty due to a lack of information/data. Accounting for both epistemic and aleatoric dynamics uncertainty in a theoretically-grounded way remains an open problem. We introduce Local Uncertainty Conformal Calibration (LUCCa), a conformal prediction-based approach that calibrates the aleatoric uncertainty estimates provided by dynamics models to generate probabilistically-valid prediction regions of the system’s state. We account for both epistemic and aleatoric uncertainty non-asymptotically, without strong assumptions about the form of the true dynamics or how it changes. The calibration is performed locally in the state-action space, leading to uncertainty estimates that are useful for planning. We validate our method by constructing probabilistically-safe plans for a double-integrator under significant changes in dynamics. Find out more: https://lnkd.in/gQejpzUm