Sanjay Thakur

About

Sanjay Thakur’s research lies at the intersection of robotics, machine learning, and uncertainty quantification, with a central focus on enabling robots to learn robustly from human demonstrations across diverse and unpredictable environments. His most cited work (18 citations) addresses a fundamental challenge in robotic control: the failure of learned policies when training and evaluation conditions differ. Thakur proposes that training from demonstrations in varied contexts can mitigate, but not eliminate, these failures, highlighting the critical need for models that account for environmental variability. In his 2019 paper (5 citations), he advances this line of inquiry by introducing Bayesian Neural Networks for uncertainty-aware learning from demonstrations, allowing robots to not only imitate actions but also express confidence in their predictions. This work is notable for its practical implications in safe and adaptive autonomy. Thakur’s contributions are particularly valuable for students and researchers working on sim-to-real transfer, imitation learning, and reliable deployment of learning-based controllers in real-world settings.

Research Focus

Key Achievements

Top Papers

1. 1
   18 citations
2. 2
   Uncertainty Aware Learning from Demonstrations in Multiple Contexts using Bayesian Neural Networks

   5 citations · 2019

Key Collaborators

Contact & Links