A. Rupam Mahmood
University of Kindu, University of Alberta, Intel (United States)
Papers
5
Total Citations
99
H-Index
4
About
A. Rupam Mahmood is a prominent researcher at the intersection of reinforcement learning (RL) and real-world robotics, whose work addresses one of the field's most persistent challenges: translating algorithmic advances from simulation into reliable physical robot systems. His highly cited 2018 paper benchmarking RL algorithms on real-world robots (46 citations) helped establish standardized evaluation practices for the research community, while his companion work on setting up RL tasks with physical hardware (24 citations) provided practical frameworks that lowered the barrier to entry for robot learning research. Mahmood has consistently pushed beyond the comfortable abstractions of simulated environments, tackling under-explored problems such as asynchronous learning in real-time settings—recognizing that the physical world does not pause for computational updates—and developing autoregressive policy representations to generate smoother, more naturalistic robot trajectories. His more recent work on distributed computing architectures for vision-based robotic learning reflects a forward-thinking awareness of the infrastructure demands real-time RL imposes. Collectively, his contributions have meaningfully advanced the practical deployment of adaptive robotic systems, making him an essential reference point for researchers bridging the sim-to-real divide.
Research Focus
Key Achievements
Top Papers
- 1Benchmarking Reinforcement Learning Algorithms on Real-World Robots46 citations · 2018
- 2Setting up a Reinforcement Learning Task with a Real-World Robot24 citations · 2018
- 3Autoregressive Policies for Continuous Control Deep Reinforcement Learning13 citations · 2019
- 4Asynchronous Reinforcement Learning for Real-Time Control of Physical Robots12 citations · 2022
- 5