Motion path editing

In this paper we provide methods that allow for path-based editing of existing motion data. We begin by exploring the concept of path as an abstraction of motion, and show how many of the common motion editing tools fail to provide proper control over this useful property. We provide a simple extension to displacement mapping methods that provide better control over the path in a manner that is easy to implement in current systems. We then extend this simple method to avoid violation of geometric constraints such as foot- skate. demonstrate how path transformations work with constraint-based approaches to provide an interactive method for altering the path of a motion. This leads to several useful applications. A path is an abstraction of the positional movement of a character. The path encodes the direction of motion, which is different from, but related to, the orientation of the character as it moves along the path. This abstraction leads to the idea of representing a motion relative to the path, allowing the path to be altered and the motion to be adjusted accordingly. The methods we present maintain the relationship between the motion and the path. This paper is organized as follows. We begin by describing an ex- ample of what our techniques are capable of and useful for (Section 2). This discussion both motivates our methods as well as discusses their relationship to existing techniques. We then introduce the ab- straction of a path for a motion (Section 3) including methods for automatically creating it. The most basic form of path transformation, presented in Section 4, can create a new motion that follows an arbitrary path and ori- ents the character appropriately. However, this transformation may damage some of the fine details in the motion such as the crispness of footplants. Better results can be obtained by using constraint processing to explicitly maintain details, as described in Section 5. The motion is continuously updated as the user drags portions of the path. Even the most sophisticated of the methods presented works interactively in all of our examples. We conclude by discussing experiments with our prototype implementation.