Jory Denny

User-Guided Motion Planning

Even the most advanced motion planning algorithms face tremendous difficulty when presented a new or severely challenging scenario. However, humans can often easily determine an approximate solution to such problems. In this work, we seek to combine the best of both worlds, the efficiency of computing systems and the intelligence of humans. Through collaboration, we have found efficient methods for robotic motion planning.

For more information

My Publications

Motion Planning on the Medial Axis

Often in robotics, we would like motions which steer clear of obstacles. This allows for safer motions for many systems. In this work, we build planning solutions on the medial-axis of the space, or the set of points equidistance to two or more obstacles. By doing this, planned paths often have high clearance. This methodology has been explored for a variety of sampling-based techniques, including probabilistic roadmaps and rapidly-exploring random trees

For more information

My Publications

Toggle PRM

Toggle PRM is a novel planning paradigm which incorporates mapping both C-free and C-obst in a coordinated fashion. During planning, when any connection attempt between two nodes of one space fails (i.e., the simple path crosses the opposite space), we retain a witness to the failure, and add it to the opposite space's roadmap, e.g., When a connection between two free nodes in the roadmap fails a witness to the failure is saved in the obstacle map. Toggle PRM is provably more efficient than uniform random sampling and experimentally is more efficient then other contemporary samplers.

For more information

My Publications

Adaptive Motion Planning

Since planning environments are complex and no single planner exists that is best for all problems, much work has been done to explore methods for selecting where and when to apply particular planners. Adaptive motion planners seek to select appropriate planners in various regions of the environment. We have applied these to common planning paradigms such as PRMs and RRTs.

For more information

My Publications

Parallel Motion Planning
(Mentoring)

In this project, we are developing parallel algorithms for motion planning applications. In our initial work, we demonstrated that PRMs are "embarrassingly parallel". In later work, we used STAPL, the Standard Template Adaptive Parallel Library, to produce a platform independent parallel implementation. Our recent work focuses on scalable framework for parallelizing sampling-based motion planning algorithms.

For more information

My Publications

Connected Component Expansion
(Mentoring)

In this project, we are developing connected component expansion algorithms which can be used in Probabilistic RoadMaps (PRMs) to improve roadmap coverage and roadmap connectivity, especially in high degree of freedom robotic systems.

My Publications

Multi-Robot Caravanning

Caravanning occurs in many places where a group of travelers must cooperate to reach a common goal. We introduce a new solution for a heterogeneous group of robots attempting to reach predefined goals. We combine leader election with following techniques to robustly traverse a set of waypoints. We have applied this to the iRobot Create platform.

For more information

My Publications

Pursuit and Evasion Techniques
(Undergraduate)

Pursuit-Evasion is a commonly studied scenario where the pursuers seek to locate the evaders, and then either attempt to capture the evaders or drive them towards a certain location. Pursuit-Evasion is applied to the catching of prey, apprehending a criminal, or herding the other agents to a desired goal location. We are able to simulate realistic behaviors in multi-level environments with heuristic-behaviors.

For more information

My Publications