Papers
7
Total Citations
65
H-Index
4
About
Arbaaz Khan is a robotics and machine learning researcher whose work sits at the intersection of graph neural networks, multi-robot systems, and autonomous navigation. His most influential contribution, "Graph Policy Gradients" (2019, 21 citations), introduced a novel algorithm that leverages graph symmetry to scale reinforcement learning policies across large numbers of homogeneous robots — elegantly sidestepping the curse of dimensionality that plagues traditional multi-agent approaches. Building on this foundation, his 2020 paper "Graph Neural Networks for Motion Planning" (15 citations) demonstrated how GNNs' permutation invariance can guide both continuous and discrete planning algorithms, opening new possibilities for robust robot navigation in complex environments. Khan has consistently addressed the challenge of safe, scalable multi-robot coordination, with his work on unlabeled multi-robot planning (13 citations) tackling goal assignment and trajectory planning in obstacle-filled workspaces through multi-agent reinforcement learning. His earlier investigations into memory-augmented neural networks for autonomous navigation (2017) reflect a long-standing interest in end-to-end learning for real-world robot deployment. Across his body of work, Khan has helped establish graph-structured learning as a principled framework for controlling robot teams at scale.
Research Focus
Key Achievements
Top Papers
- 1Graph Policy Gradients for Large Scale Robot Control21 citations · 2019
- 2Graph Neural Networks for Motion Planning15 citations · 2020
- 3Learning Safe Unlabeled Multi-Robot Planning with Motion Constraints13 citations · 2019
- 4Neural Network Memory Architectures for Autonomous Robot Navigation5 citations · 2017
- 5
- 6End-to-End Navigation in Unknown Environments using Neural Networks4 citations · 2017
- 7Sufficiently Accurate Model Learning3 citations · 2020