Reinforcement Learning and World Models for Robotics

​Robotics is steadily evolving, and two ideas, reinforcement learning and world models, are increasingly shaping how we build more capable and adaptive systems. This session will walk through the current state of robotics, how learning based methods are complementing or replacing traditional control strategies, and what it might take for robots to handle the complexity of real world environments more reliably.

​We’ll touch upon:

• The limitations of classical control pipelines
• Where and how learning-based systems are making a difference
• Techniques like learned simulators, reward shaping, and closed-loop training
• The open challenges on the path to deploying general-purpose robots beyond the lab

Speaker:
Mankaran Singh is the founder of the FlowDrive project and is currently building in the consumer robotics space. He previously worked at Addverb, Ola Electric, and the Indy Autonomous Challenge. His work spans industrial automation, electric mobility, and autonomous systems.
LinkedIn: /mankaran32
X: /Mankaran32