IISc HiRo Lab

The Human-interactive Robotics (HiRo) Lab is a research group in the Robert Bosch Centre for Cyber-Physical Systems at the Indian Institute of Science Bengaluru set up and led by Prof. Ravi Prakash.

Our group’s research interests lie at the intersection of robotics, machine learning and human-robot interaction - with a focus on robot manipulation for real world applications i.e., healthcare, retail, space robots etc. The central principle is to enable humans with different levels of robotic expertise to transfer their knowledge and experience about skills and tasks to the robot to explore and execute the tasks in unstructured novel scenarios.

Research advances (sketched above) are targeted in the domain of <!–

• Interactive Skill Learning
  • Imitation Learning via Dynamical Systems, DMPs, Gaussian Processes
  • Imperfect Human Demonstration
  • Uncertainity Estimation
  • Bimanual Manipulation
• Generalization in New Situations
  • Generalization of Policy via Task-Parameterization (i.e., frame, object, obstacle, landmarks)
  • Generalization of Task via Object Relationship Modeling (i.e., affordances)
• Long Horizon Task
  • Hierarchical Reinforcement Learning
  • Task and Motion Planning/Learning
  • Spatio-Temporal Specification for Time-Critical Scenarios
• Low Level Robot Control
  • Impedance Control for HRI
  • AI augmented Optimal Control
  • Stability Guarentee for Interconnected Systems
  • Fail Safe Motion in Human Environment –>
• Toward a Science of Data for Robot Learning
  • Establishing principles for the composition and curation of multimodal learning datasets
  • distinguishing useful variability from noise
  • and enabling robots to learn effectively from mixed-quality demonstrations.
• Generalization and Robustness under Distributional Shifts
  • Generalization of policy via Task-Parameterization (i.e., frame, object, obstacle, landmarks)
  • Developing continual learning controllers with certified stability that adapt safely to evolving task conditions
  • Generalization of task via object relationship modeling (i.e., affordances)
• Safe and Reliable Control of Autonomous Systems
  • Impedance control for HRI
  • AI augmented Optimal Control and Reinforcement Learning
  • Advancing modular safety filters and decentralized safety mechanisms for cooperative, contact-rich, or human-in-the-loop systems.
• Language, Agents, and World Models
  • Building embodied foundation models that unify language, perception, and control
  • Diffusion-based world modeling and reasoning-guided policy learning.