Hi there! I'm a fourth year PhD student here at Berkeley, working on autonomous robotics and optimal control. I am advised by Prof. Claire Tomlin in the Hybrid Systems Lab. I'm also a member of the Berkeley AI Research lab, and I am supported by an NSF Graduate Research Fellowship. I did my undergrad in electrical engineering at Princeton (2015).
Outside of research, I like to play squash and frisbee, read fantasy novels, and playing acoustic guitar.
The best way to reach me is by email. My address is: dfk at eecs dot berkeley dot edu.
I am generally interested in optimal control, motion planning, and safe autonomy. So far, I have primarily worked on techniques for safe motion planning that provide robustness guarantees and operate in real-time. My work so far falls into two main categories: efficient motion planning with reachability-based safety guarantees, and robust adaptive control with a priori unknown dynamics. I have also worked on a number of other projects related to distributed control, adaptive receding horizon control, and active search.
In motion planning and control, there is often a division between methods that work well in real time, and those that provide strict safety guarantees. For example, iterative LQR (iLQR) is a popular method of generating smooth optimal control sequences in real time for relatively high-dimensional systems; yet iLQR does not guarantee robustness, e.g. against environmental disturbances. By contrast, Hamilton-Jacobi (HJ) reachability can provide hard safety guarantees and disturbance rejection but for general nonlinear systems is only tractable in fairly low dimension.
FaSTrack is a new approach that uses an offline reachability computation to inform an online motion planner like iLQR to blend the best of these two types of algorithms. I have worked on several projects extending FaSTrack. In Planning, Fast and Slow, we broaden the concept of motion planning to allow for multiple different types of planning algorithms (with potentially different notions of state) to run concurrently and stitch together seamlessly while retaining the original FaSTrack safety guarantee. We took a first step toward incorporating uncertain predictions of a pedestrian into FaSTrack in a recent paper at RSS 2018, and we recently extended this approach to work with multiple pedestrians and multiple robots. We have also tested a neural network-based HJ reachability solver and used it to compute conservative approximations to reachable sets for higher dimensional systems. Finally, I recently built a high-level graph-based wrapper around kinodynamic planners to extend the modular FaSTrack guarantees to a priori unknown environments while retaining recursive safety and liveness.
The capability of handling some level of uncertainty in system dynamics is one feature of HJ reachability that makes it attractive for robust control. However, the characterization of uncertainty is typically represented by additive disturbances in state variables, rather than by uncertainty in the parameters of the dynamics, or by stochasticity. I am currently working on a new strategy for this regime based on minimax approaches to robust control.
While there is a lot of good work going on in motion planning, control, prediction, perception, etc., it strikes me that we still lack a formal understanding of the closed loop system
dynamics that emerge when these modules are all connected. Specifically, much of my prior work has focused on the interaction between motion planning and lower-level control and reference tracking. In this setting, we can find a
separation principle between planning and control that allows a very modular design, yet preserves strong real-time robust, operational safety guarantees.
I am interested in trying to extend these methods to include uncertain interactions with other agents in the environment. For example, in autonomous driving, it is almost never possible to know exactly how other vehicles will behave in the future; yet, we can often make reasonable guesses. It would be incredibly useful if there were a structured way to co-design a motion planner and a behavior predictor so that, together, they guarantee perpetual collision-avoidance.