# USC Brings the World’s Control and AI Researchers to Campus – Together > Source: > Published: 2026-06-29 17:46:56+00:00 Picture a Waymo stuck in a four-way standoff, unable to decide who goes first. It’s a small glitch with a simple cause: the car’s AI is great at learning patterns, but less effective at reasoning through novel situations with any guarantee it will get them right. That gap between systems that learn and systems we can actually trust is what brought hundreds of researchers to the [USC Viterbi School of Engineering](https://viterbischool.usc.edu/) and the [USC Stevens School of Advanced Computing and AI](https://stevens-computing-ai.usc.edu/) this June for two conferences happening at once. From June 16-19, researchers from around the world gathered at USC for the [Learning for Dynamics & Control Conference (L4DC)](https://sites.google.com/usc.edu/l4dc2026/) and the [International Conference on Neuro-Symbolic Systems (NeuS)](https://sites.google.com/usc.edu/neus2026/home), two events exploring different approaches to the same challenge. “Today, the dynamics and control communities are working closely with the machine learning community, and there is considerable convergence of interest in neuro-symbolic systems across these disciplines. We felt it would benefit all attendees of both meetings to learn from each other, so co-locating these two meetings made a lot of sense,” said [Gaurav Sukhatme](https://uscresl.org/principal-investigator/), USC Viterbi executive vice dean, inaugural director of the USC Stevens, and general chair of L4DC 2026. ## Two Paths, Same Goal While both conferences were concerned with building more reliable intelligent systems, they focused on different pieces of the problem. **L4DC focused on learning and control.** Researchers explored how intelligent systems interact with dynamic environments, from autonomous vehicles and robotics to aerospace systems and energy networks. “L4DC is an interdisciplinary conference for researchers focused on making learning-enabled feedback systems act reliably in the real world,” agreed [Stephen Tu](https://stephentu.github.io/), Lars Lindemann, Nikolay Atasanov and Adam Wierman, who served as the program co-chairs for L4DC. The conference program co-chairs themselves span multiple facets of the L4DC community. Tu’s interests are in learning and control for dynamical systems, generative modeling, and robotics. He is with the [Ming Hsieh Department of Electrical and Computer Engineering](https://minghsiehece.usc.edu/) at USC. Lindemann, at ETH Zurich, specializes in systems and control theory, formal methods, and machine learning. Atanasov’s research focuses on reliable, efficient, and versatile autonomous robotic systems – he is with the Electrical and Computer Engineering department at UCSD. Wierman, an expert in resilient networked systems, is with the Computing and Mathematical Sciences department at Caltech. **NeuS focused on learning and reasoning.** Researchers combined machine learning with symbolic reasoning and formal methods, bringing together data-driven AI and more structured approaches to decision-making. One example is autonomous driving. Machine learning systems excel at navigating complex and unpredictable environments, but researchers also want stronger guarantees about how those systems will behave. Together, the conferences explored three interconnected areas of AI research: learning, control and reasoning. More broadly, they explored a common challenge: building systems that can both make decisions and carry them out safely in the real world. For Alessandro Abate, keynote speaker at L4DC 2026 and professor at the University of Oxford, those aren’t separate fields. “I work at the intersection of these three areas,” he said. “I find it very natural for these two events to be collocated.” ## A Shared Conversation The conferences unfolded over four days across the [USC Michelson Center for Convergent Bioscience](https://michelson.usc.edu/) and [Dr. Allen and Charlotte Ginsburg Human-Centered Computation Hall](https://viterbischool.usc.edu/ginsburghall/), with attendees moving between talks, tutorials, poster presentations and discussions throughout the week. The arrangement reflected the growing overlap between the two communities. Researchers studying learning and control could attend sessions on reasoning and formal methods, while NeuS attendees could explore how those ideas are being applied to robotics, autonomous systems and other real-world challenges. The two events also shared spaces, meals and social gatherings, allowing conversations sparked in one conference to carry into the other. Ideas moved between conferences as readily as attendees did, from keynote talks to coffee-break conversations. ## From Theory to Practice For researchers working in learning and control, the goal is not simply building more capable algorithms but creating systems that can function safely and effectively in the real world. “If you don’t want to be stuck in a deadlock of Waymo vehicles at an intersection, then you need to value learning for dynamics and control,” said [Ketan Savla](https://viterbi.usc.edu/directory/faculty/Savla/Ketan), an L4DC organizer and the [John and Dorothy Shea Early Career Chair in Civil Engineering](https://cee.usc.edu/) at the USC Viterbi School. This year’s L4DC introduced new industry-focused programming, including spotlight talks and a panel discussion featuring leaders from industrial research labs, aerospace, autonomous trucking and commercial robotics. According to Tu, the additions were intended to help ensure research remains connected to problems that matter in practice. ## Teaching Machines to Reason “Learning-based approaches can do a lot of amazing things, but they can’t really guarantee anything.” That’s how Armando Solar-Lezama, MIT professor and program co-chair of the International Conference on Neuro-Symbolic Systems (NeuS), described one of the biggest challenges facing artificial intelligence today. While L4DC focused on how intelligent systems interact with the world, NeuS explored how they reason and can be verified within it, helping AI move beyond recognizing patterns to understanding context, relationships and rules. “Formal reasoning is fundamental to artificial intelligence, as is verifiability and proven reliability,” said [Jyotirmoy Deshmukh](https://jdeshmukh.github.io/), general chair of NeuS 2026 and associate professor in the [Thomas Lord Department of Computer Science](https://www.cs.usc.edu/) with a joint appointment in [USC’s Ming Hsieh Department of Electrical and Computer Engineering](https://minghsiehece.usc.edu/). [Anura Deshpande](https://www.linkedin.com/in/anura-deshpande/), a USC senior studying computer science and linguistics, attended NeuS sessions and tutorials as a student volunteer. One presentation that stood out explored how robots reason about objects and their surroundings. “To explain to a robot how to pick up a certain object, it has to know what’s around that object as well,” Deshpande said. For Deshpande, it was a useful example of the kinds of reasoning problems researchers at NeuS are trying to solve. The conference also highlighted how ideas developed in one field often find applications in another. Nicholas K., a master’s student at UC San Diego attending his first research conference, said he was struck by the similarities between his own work and research being presented in other disciplines. “My research right now is on human motion generation for sports,” he said. “It’s cool to hear these concepts being talked about in completely different applications.” Published on June 29th, 2026 Last updated on June 29th, 2026