Loading Events

« All Events

  • This event has passed.

ESE Ph.D. Thesis Defense: “Graph Convolutions for Teams of Robots”

November 18 at 12:00 PM - 1:30 PM

In many applications in robotics, there exist teams of robots operating in dynamic environments requiring the design of complex communication and control schemes. The problem is made easier if one assumes the presence of an oracle that has instantaneous access to states of all entities in the environment and can communicate simultaneously without any loss. However, such an assumption is unrealistic especially when there exist a large number of robots. More specifically, we are interested in decentralized control policies for teams of robots using only local communication and sensory information to achieve high-level team objectives. We first make the case for using distributed reinforcement learning to learn local behaviors by optimizing for a sparse team-wide reward as opposed to existing model-based methods. A central caveat of learning policies using model-free reinforcement learning is the lack of scalability. To achieve large-scale scalable results, we introduce a novel paradigm where the policies are parametrized by graph convolutions. Additionally, we also develop new methodologies to train these policies and derive technical insights into their behaviors. Building upon these, we design perception-action loops for teams of robots that rely only on noisy visual sensors, a learned history state, and local information from nearby robots to achieve complex team wide-objectives. We demonstrate the effectiveness of our methods on several large-scale multi-robot tasks.

Arbaaz Khan

ESE Ph.D. Candidate

Arbaaz Khan is a PhD candidate at the GRASP Lab at University of Pennsylvania where I am advised by Vijay Kumar and Alejandro Ribeiro. His research interests lie at the intersection of machine learning and robotics. His research interests lie at the intersection of learning and decision making for robotics. These include improving upon existing policy optimization methodologies, neural network architectures and exploring new models for robust learning. Most recently he has been investigating the use of graph convolutions for problems in robot path planning and multi-robot control. Before his PhD, he was pursuing a masters in Robotics also at the University of Pennsylvania where he was advised by Professors Daniel D. Lee and Vijay Kumar. During his masters, his research was focused on learning meaningful representations for robot navigation. He also spent some time working at the LAIR Lab at Carnegie Mellon University’s Robotics Institute. At CMU, his research was focused on autonomous UAV flight through GPS denied cluttered outdoor environments such as forests. He has also been fortunate to spend time as a research intern at NVIDIA, Samsung Research, Apple and Google Brain.


November 18
12:00 PM - 1:30 PM


Electrical and Systems Engineering


Room 452 C, 3401 Walnut
3401 Walnut Street
Philadelphia, PA 19104 United States
+ Google Map