Anticipate, Adapt, Act: A Hybrid Framework for Task Planning

Nabanita Dash¹    Ayush Kaura¹    Shivam Singh¹    Ramandeep Singh¹    Snehasis Banerjee²    Mohan Sridharan³    K Madhava Krishna¹   

¹ Robotics Research Center, IIIT Hyderabad, India    ² TCS Research, Tata Consultancy Services, India    ³ School of Informatics, University of Edinburgh, UK   

Anticipating and adapting to potential failures is a key capability that robots require for effective human–robot collaboration in complex domains. Despite advances in AI planning systems and Large Language Models (LLMs), uncertainty in task outcomes remains a challenge. We introduce a hybrid framework that integrates LLM predictions with relational decision-making using RDDL. The robot reasons about human capabilities, possible task failures, and executes actions to either prevent or recover from failure. In experiments (VirtualHome 3D simulation), our method shows improvements in task completion, execution time, and collaboration.