My research aims to build a unified algorithmic framework where a robot efficiently infers the optimal value function from a bounded set of interactions with both humans and the environment. It ties together insights from motion planning and imitation learning and applies them to robots deployed in the wild.

2018 - 2019
| Generalized Lazy Search Typical search algorithms expand a wavefront from the start, evaluating edges discovered until the shortest path is found. Hence planning time is a sum of |

2016 - 2017
| Incremental Densification The gold standard in motion planning is to find the shortest path in continous space. One way to achieve that is to create a dense fully-connected graph and search it, however, that will take O(N |

2015 - 2016
| Hybrid Local and Global Search Sampling-based planners like BIT* are effective at exploring the whole solution space but are slow to exploit. On the other hand, trajectory optimizers like CHOMP locally exploit around an initial path but running it on all potential paths is too expensive. Our key insight is to |

2012 - 2013
| Planning Alternate Routes using a Sampling Based Planner We address the problem of finding multiple |

2013 - 2014
| Speedy Search on Sparse 3D Visibility Roamdaps We have designed a 3D planning algorithm, SPARTAN, that is orders of magnitude faster than sampling-based or discrete search alternatives. We do so by constructing the sparsest possible roadmap that contains the shortest 3D path. Our key insight is that the shortest path is a |

2014 - 2015
| Fast Nonlinear Trajectory Optimization using Filtering Techniques Nonlinear trajectory optimization for mobile robots has two main components - minimizing an objecting and enforcing dynamic constraints. The objective is often defined only on the 3D workspace (e.g. avoid obstacles, maximize smoothness) while constraints are on the full state space. Our key insight is to first optimize a workspace trajectory and then |

2016 - 2017
| Trajectory Optimization in a Moving Frame We address the problem of planning long, dynamically feasible, time-optimal trajectories in the presence of wind (which creates a moving reference frame). We present an algorithm, KITE, that elegantly decouples the joint trajectory optimization problem into individual path optimization in a fixed ground frame and a velocity profile optimization in a moving reference frame. |

2015 - 2016
| Optimal Repairing of Vector Fields This paper presents a framework that integrates vector field based motion planning techniques with an optimal path planner. Our framework uses a vector field as a high level specification of a task and an optimal motion planner (in our case RRT*) as a local, on-line planner that generates paths that follow the vector field, but also consider the new obstacles encountered by the vehicle during the flight. |