We combine rigorous mathematical foundations with hardware validation in laboratory and field environments, advancing the frontier of autonomy in space exploration.
CADRE (Cooperative Autonomous Distributed Robotic Exploration) is a lunar technology demonstration of multi-agent autonomy on a team of three rovers and a base station. Receiving only high-level tasks from Earth, the rovers autonomously explore a region of the lunar surface and perform a distributed measurement in coordination with a multi-static ground-penetrating radar.
Ice worlds are at the forefront of astrobiological interest because of evidence of subsurface oceans. To sample Enceladus' vent systems, NASA JPL is developing a snake-like robot, EELS, that navigates extreme surfaces and descends erupting vents to capture unaltered liquid samples — potentially reaching the ocean. We contribute dynamics modeling and resilient autonomy for such platforms.
The DARPA Learning Introspective Control (LINC) program develops machine-learning-based introspection that enables physical systems — ground vehicles, ships, drone swarms, robots — to respond to events not predicted at design time. LINC updates control laws in real time while providing guidance and situational awareness to the operator, human or autonomous.
Inspection and mapping of a target spacecraft in low Earth orbit using multiple observer spacecraft in stable passive relative orbits is a key enabling technology for future missions. Our guidance and control architecture uses an information-gain approach to directly trade off gathered data against fuel and energy cost.
We developed generalized polynomial chaos-based sequential convex programming (gPC-SCP) to compute suboptimal solutions for continuous-time chance-constrained stochastic nonlinear optimal control problems, enabling motion planning under uncertainty. The method first derives a deterministic surrogate with convex constraints via gPC expansion and a distributionally robust convex subset of the chance constraints, then solves it using sequential convex programming for trajectory generation and control.