Maaz.

Research agenda · Autonomous systems

Questions before affiliations.

I study how autonomous systems can make reliable decisions, coordinate at scale, and learn without discarding the structure that makes control systems trustworthy.

01

Decision Making Under Uncertainty

How should machines make decisions before they know enough?

Stochastic trajectory familyState / time
π* = arg min 𝔼 [Σ ℓ(xₜ,uₜ) + ℓf(xT)]
x₀ TARGET SET

Methods

Stochastic control · MPC · MPPI · DDP · reinforcement learning · trajectory optimization

Systems

Aircraft · UAVs · robotic systems

Recent work

Autonomous eVTOL Control →AI Grand Prix →

02

Scalable Multi-Agent Optimization

How can many autonomous agents coordinate when centralization stops scaling?

Methods

Distributed optimization · deep unfolding · decomposition · coordination policies

Systems

Swarm systems · air traffic · resource allocation · heterogeneous fleets

Current direction

UnfoldAI

Distributed coordination graphIteration k
123 456 789
Primal residual10⁰10⁻⁴

03

Learning + Control for Physical Systems

How can learned models augment control without giving up structure?

Hierarchical learning-control architectureClosed loop
  • Target
  • Learning + MPC
  • Unconstrained policy

Methods

RL-MPC · learned control · robust control · sim-to-real · state estimation

Systems

Bipedal locomotion · eVTOL autonomy · perception-driven robotics

Recent work

Autonomous eVTOL Control →Bipedal Locomotion →