The motion planning group at UNC has been working on various planning problems. Some of the earlier work dealt with using Voronoi regions and medial axis to improve the performance of probabilistic approaches. More recently, we have been working on constrained-based planning, deformable models and complete motion planning algorithms for rigid robots.

   We have implemented several distinct motion planning systems:

• Complete Motion Planning
  
• Physics-based Motion Planning (Deformable robots)
  
• Composite Agents (crowd and multi-agent simulation)
  
• Virtualized Traffic (Traffic Reconstruction)
  
• Independent Navigation of Multiple Mobile Robots (mobile robot navigation)
  
• Multi-robot coordination and planning (multi-robot coordination)
  
• Constrained-based Motion Synthesis for Deformable Models (deformable models)
  
• D-Plan: Efficient collision-free path computation for part removal and disassembly (CAD/CAM applications)
  
• Virtual Agents for Online Virtual Worlds (Virtual Worlds and Second Life)
  
• Real-time Navigation of Independent Agents using Adaptive Roadmaps (crowd and multi-agent simulation)
  
• Reciprocal Velocity Obstacles for Real-Time Multi-Agent Collision Avoidance (multiple robot planning)
  
• Interactive Navigation of Individual Agents in Crowded Environments (crowd and multi-agent simulation)
  
• Reactive Deforming Roadmaps (multiple robot planning)
  
• Real-time Path Planning for Virtual Agents in Dynamic Environments (crowd and multi-agent simulation)
  
• An Efficient Retraction-based RRT Planner (handling narrow passages)
  
• Star-shaped Roadmaps - A Deterministic Sampling Approach for Complete Motion Planning:
  
• Path Planning for Deformable Robots in Complex Environments (flexible objects)
  
• Practical Local Planning in the Contact Space:
  
• Constraint-Based Motion Planning of Deformable Robots:
  This planning framework is an extension to Constraint-Based Motion Planning. We generalize a list of constraints that are suitable for robots that can bend and flex in a physically realistic manner. The deformable robot travels along a guiding path while the planner makes appropriate path adjustments and corrections to compute a collision-free path.
• Interactive Navigation in Complex Environments Using Path Planning:
  This system uses path planning techniques to allow a user to navigate an avatar through extremely large and complex virtual environments. As a preprocess, a randomized sampling technique constructs a graph representing the environment and prunes the graph to maintain only  those nodes that are useful for navigation. At runtime, the user can specify a start and goal location and have the path between them computed in real-time.
• Real-Time Constraint-Based Motion Planning:
  This planning framework transforms the problem of motion planning into constrained dynamics simulation. It is applicable to dynamic environments with multiple moving obstacles and collaborating, articulated agents, subject to various type of geometric constraints. It's extensible to include other types of constraints and deformable objects.
  
• V-Plan:
  This is a hybrid planner that generates an estimated path using the medial axis of the workspace, and then corrects it using localized randomized planning. This planner is an example implementation within a framework that we have proposed.
  
• A Randomized Path Planner:
  This planner uses the medial axis primarily to bias the generation of random samples.
  
• A Real-Time 3-dof Potential Field Planner:
  This planner handles dynamic scenes by using Voronoi-guided potential field methods.

• Our randomized planning software relies on collision detection using the package PQP, developed here at UNC.
• We have tested our software on models provided by the Move3D project at LAAS, Toulouse, and the DSMFT group at Texas A&M University.
• We use randomized motion planning software written by David Hsu while a student at Stanford.