Planning Algorithms for Computer Animation: from Humanlike Search Spaces to Local Clearance Triangulations
I will present in this talk our latest results for addressing the problem of humanlike motion planning. Our starting point is a multi-modal planning framework that coordinates locomotion, body positioning, and upper-body action execution according to coordination patterns extracted from human subjects. Blending spaces defined from example motions are then introduced for achieving continuous search spaces describing humanlike variations of generic actions. The approach is complemented with a virtual reality interface for modeling blending spaces from direct demonstration. As a result, humanoid assistants or autonomous virtual characters are able to plan and execute motions that are similar to demonstrated examples, and at the same time addressing new constraints and parameterizations. I will also present new results for the efficient computation of navigation queries from local clearance triangulations, a new structure I have developed for fast path planning with arbitrary clearance. The approach is suitable for handling large and non-static environments, and is being adopted by the computer game industry.
Marcelo Kallmann is founding faculty and associate professor of computer science at the University of California, Merced. Before moving to UC Merced he was a research faculty member at the computer science department of the University of Southern California (USC) and a computer scientist at the USC Institute for Creative Technologies (ICT). Between 2001 and 2004 he did postdocs at the USC robotics research lab and at the virtual reality lab of the Swiss Federal Institute of Technology (EPFL), where he completed his PhD in 2001. His areas of research include computer animation, virtual reality, motion planning and humanoid robotics. He routinely serves as a program committee member for major conferences in these areas and in 2012 he will be program co-chair for the International Conference on Motion in Games (MIG).