- Formal methods + controls: The methods we have developed address automated synthesis of control protocols that rely on integration of physical laws and software principles to serve in adversarial environments subject to rich temporal-logic-like specifications.
- Learning + formal methods: The central question is how we can develop autonomy protocols that not only learn from their interactions with the environment and users but also provably satisfy high-level safety and performance specifications.
- Controls + learning: The main question is how we can guarantee safety and robustness feedback control systems that integrate learning-enable components, e.g., classifiers, in the loop.
In addition to the applications in aerial and ground vehicles (and robots), we interpret autonomy broadly with other emerging applications in networks on large-scale aerospace systems and additive manufacturing.
We create our particular projects often by abstracting problems from one or more pressing needs in autonomous systems. The current needs my group’s work addresses include (i) the verifiability of autonomous systems; (ii) their adaptability using data at design- and run-time; (iii) their co-work with human users; (iv) the explainability of autonomous decisions, potential failure reasons and necessary fixes to users and designers; and (v) the limitations faced by autonomous systems due to imperfect sensing and perception.
- Data-Driven Cyberphysical Systems (NSF, CPS)
- CAREER: Provably Correct Shared Control for Human-Embedded Autonomous Systems (NSF)
- Hybridizing Learning and Model-Based Planning for Active Perception (ONR)
- Explainable and Scalable Planning with Probabilistic Temporal Logic Specifications (NASA)
- Exploiting Symmetries in Software for More Robust and Efficient Systems (DARPA)
- Compositional Verification of Hybrid Systems (AFRL, RQ)
- Autonomous Detection and Assessment with Moving Sensors (Sandia National Labs)
- EAGER: Human-Aware Navigation in Populated Indoor Environments (NSF, NRI)
- High-Confidence, Efficient Learning Under Rich Task Specifications (NSF, RI)
- Probably Approximately Correct Protocols for Reactive Control and Learning (ARO)
- Formal Synthesis of Collaborative Protocols for Joint Learning and Control (DARPA)
- Formal Specification and Correct-by-Construction Synthesis of Control Protocols for Adaptable, Human-Embedded Autonomous Systems (AFRL, RQ)
- Risk-Aware, Human-Cooperative Planning for Autonomous Systems (ONR, subcontract from JPL)
- Synthesis of Correct-By-Construction Control Protocols for Open, Reconfigurable Shipboard Networks (ONR)
- Autonomy Protocols: From Human Behavioral Modeling to Correct-By-Construction, Scalable Control (NSF, CPS)
- STTR (Phases I and II): Correct-by-Construction Synthesis for Multi-Vehicle Autonomy Missions (AFOSR, subcontract from Galois Inc.)
- Perception-Based, Reactive, Temporal-Logic Planning For Autonomous Deck Operations (ONR)
- Specification, Synthesis and Verification of Software-Based Control Protocols for Fault-Tolerant Space Systems (AFRL, RV)
- Architectural and Algorithmic Solutions for Large-Scale PEV Integration into Power Grids (NSF, CPS)
- Formal Synthesis of Software-Based Control Protocols for Fractionated, Composable Autonomous Systems (AFOSR, YIP)