Muhammad Aneeq uz Zaman

Muhammad Aneeq uz Zaman

PhD student

University of Illinois Urbana Champaign

I am a PhD candidate under the advisorship of Professor Tamer Başar at the University of Illinois at Urbana Champaign. I did my Master’s from the University of Illinois at Urbana Champaign under the advisership of Professor Naira Hovakimyan in 2015.

My research focuses on Multi-agent Reinforcement Learning (MARL) using the paradigm of Mean-Field Games (MFGs). I am primarily interested in learning macro-scale phenomena that emerges from micro-scale inter-agent strategic interactions in real-world scenarios, such as, consensus formation over networks, peer-to-peer power markets, adversarial influence on opinion dynamics etc. I am quite excited about Mean-Field Games as a viable method to understand and analyze these interactions.

Download my resumé .

Interests
  • Multi-agent Reinforcement Learning (MARL)
  • Reinforcement Learning for Mean-Field Games (MFGs)
  • Reinforcement Learning for Mean-Field Type Games (MFTGs)
  • Age of Information in MFGs
  • Peer-to-peer power markets
Education
  • PhD candidate in Mechanical Engineering, 2018-23

    University of Illinois Urbana Champaign

  • M.S. in Mechanical Engineering, 2013-15

    University of Illinois Urbana Champaign

  • B.E. in Mechatronics Engineering, 2006-10

    National University of Sciences and Technology, Pakistan

Skills

Python (Pandas, TensorFlow)
C++
MATLAB/Simulink
Mathematica
ROS
PCL
X-Plane
AutoQuad
Player/Stage

Experience

 
 
 
 
 
Research Assistant
Jan 2018 – Present Urbana, Illinois

Responsibilities include:

  • Reinforcement Learning for LQ Mean-Field Games (MFGs)
  • Reinforcement Learning for LQ MFGs in multiple populations
  • Oracle-free Reinforcement Learning for MFGs
  • Reinforcement Learning for Measn-Field Type Games (MFTGs)
 
 
 
 
 
Summer Intern
May 2018 – Aug 2018 Warren, Michigan

Responsibilities include:

  • I designed, implemented and tested fault diagnosis, failure prediction modeling and Remaining Useful Life (RUL) estimation for the battery systems of electric vehicles. A combination of model based and data driven techniques were used to obtain the modules using real world data.
 
 
 
 
 
Lecturer
Aug 2015 – Jul 2017 Islamabad, Pakistan

Responsibilities include:

  • As a lecturer I taught Solid Modeling, Numerical Methods, Electro-Mechanical Systems and Control Systems Design.
 
 
 
 
 
Research Assistant
Jun 2015 – Aug 2015 Urbana, Illinois

Responsibilities include:

  • During the research assistantship I worked on sensor fusion algorithms, adaptive control and autopilot-base station communication using MavLink messages.
 
 
 
 
 
Summer Intern
Feb 2015 – May 2015 Urbana, Illinois

Responsibilities include:

  • I was involved with developing the mathematicalmodel of the propeller-motor assembly of the DJI F550. Also worked on emulating the F550 in the X-Plane environment and mass properties calculator.
 
 
 
 
 
Design Engineer
Public Sector Organization
May 2011 – Dec 2012 Islamabad, Pakistan

Responsibilities include:

  • I worked on design and fabrication of Electro-mechanical Actuator Amplifiers and their testing jigs.
 
 
 
 
 
Design Engineer
Public Sector Organization
May 2011 – Dec 2012 Islamabad, Pakistan

Responsibilities include:

  • I worked on design and fabrication of Electro-mechanical Actuator Amplifiers and their testing jigs.
 
 
 
 
 
Applications and Design Engineer
Dec 2010 – May 2011 Lahore, Pakistan

Responsibilities include:

  • I was involved with competitor study and developing new features for the IntelliMax DCS software.

Projects

*
Age of Information for Multi-agent systems
Design ‘optimal’ policies for agents interacting over a communication constrained networks using Age of Information (AoI) metric
Oracle-free Multi-agent RL
Design ‘optimal’ policies for agents interacting over a multi-graphs (physical and social networks) in the presence of adversaries
Multi-Agent RL in multiple populations
Design and analysis of RL algorithms for multi-agents in multiple populations, through the lens of Mean-Field Games
Safety and Security of Multi-agent Systems
Design ‘optimal’ policies for agents interacting over a multi-graphs (physical and social networks) in the presence of adversaries
Optimal path planning for thermalling gliders
We propose an algorithm to plan shortest paths (provably) for gliders to visit a set of waypoints in the presence of thermals (refueling points)

Recent Publications

#Quickly discover relevant content by filtering publications.

(2022). Weighted Age of Information based Scheduling for Large Population Games on Networks. Weighted Age of Information based Scheduling for Large Population Games on Networks.

PDF Cite Project Slides Custom Link

(2022). Oracle-free Reinforcement Learning in Mean-Field Games along a Single Sample Path. Oracle-free Reinforcement Learning in Mean-Field Games along a Single Sample Path.

PDF Cite Project Slides Custom Link

Accomplish­ments

Coursera
Welcome to Game Theory
See certificate
Coursera
Algorithms on Graphs
See certificate
Coursera
Neural Networks and Deep Learning
See certificate

Contact