Robot Talk Episode 166: Rika Antonova on Evolving Robot Bodies and Brains

Robot Talk episode 166 presents a conversation between host Claire and Rika Antonova, an associate professor at the University of Cambridge. The episode focuses on how artificial intelligence can be used to develop robot hardware and algorithms that work well together.

A central theme is co-design, an approach in which a robot's physical body and its control algorithms are developed in tandem. Instead of fixing the hardware first and then writing software for it, co-design allows the morphology and the brain to influence each other throughout the design process.

Antonova's research interests include co-design, automated algorithm discovery, data-efficient reinforcement learning, exploration, active learning, and robotics. These areas connect to the challenge of creating robots that can learn from limited experience and adapt to new tasks.

Data-efficient reinforcement learning is particularly important in robotics because real-world interaction is often slow, costly, or risky. Methods that learn from fewer trials can make robot training more practical.

Active learning and exploration give a robot some control over the data it collects. Rather than passively receiving examples, the robot can choose actions or queries that are most informative for improving its behavior.

Automated algorithm discovery uses computational search to find algorithms or controllers, potentially reducing reliance on manual engineering. When combined with co-design, it can help identify software that is well suited to a specific robot body.

The episode's title, "Evolving robot bodies and brains," reflects this intersection. Evolutionary and learning-based methods can search over both physical designs and control policies, though the podcast discussion remains accessible to a general audience.

Robot Talk is a podcast series that interviews robotics researchers about their work. Episode 166 continues that format, offering listeners an overview of how AI might contribute to more integrated robot design.

Antonova's dual background in academia and industry informs her perspective. She has worked at Google as a senior software engineer, and her research career includes positions at leading institutions.

By bringing together hardware and algorithms, the research discussed in the episode points toward robots that are not only more capable but also more efficient to develop. The conversation highlights ongoing efforts to make robot design more automated and data-driven.

The podcast format allows researchers to explain complex topics without requiring listeners to have specialized knowledge. Claire's questions guide the conversation through key concepts and examples.

For those interested in embodied AI, the episode provides an accessible entry point to co-design and learning-based control. It shows how AI can be a partner in both hardware and algorithm development.