Learning, control and decision-making for the physical world.
We develop algorithms and engineered systems that connect learning, decision-making and the physical world. Our research spans trustworthy autonomy, planning under uncertainty, intelligent communications, digital fabrication and privacy-preserving sensing.
ConsDRED combines distributional reinforcement learning with stochastic model predictive control to track trajectories while respecting constraints.
Probabilistic constrained reinforcement learning with formal interpretability, evaluated in real quadrotor flight.
Inducing and detecting neural modules, then testing their functional roles through interventions on learned policy networks.
Mission-level routing, resource allocation and adaptive replanning as operating conditions change.
Research in European Journal of Operational Research and IEEE Internet of Things Journal.
Explore planning research ↗Multi-agent learning, aerial networks and reconfigurable wireless systems, with an emphasis on performance, fairness and explainability.
Research in IEEE TMLCN, Ad Hoc Networks and IEEE OJ-COMS.
Explore communications research ↗
Designing and fabricating functional electronics on curved and freeform surfaces—from thermoformed circuits to conformal antennas and multi-layer power electronics. Open5x and MM-Open5x provide open platforms for multi-axis, multi-material printing.
Explore fabrication research ↗Understanding and limiting unintended inference from sensor data, including paralinguistic information carried by speech.
Research published in ACM TOPS, Interspeech and CCW.
Explore sensing research ↗