A Motion Capture based Planner for Virtual Characters Navigating in 3D Environments

Abstract

In this work, a strategy to automatically generate eye-believable motions for a virtual character that navigates in a 3D environment is presented. The overall approach consists of four components as follows. (1) A state-of-the-art path planner that computes a collision-free reference path for the character’s center of mass (COM). For this planner, a simplified model that bounds the character’s geometry is proposed. (2) A segmentation algorithm that divides the path into behaviors. (3) A classifier that compares each behavior with the corresponding motion capture segments previously analyzed and stored in a database. (4) A whole-body motion generator that synthesizes the appropriate behavior determined by the classifier. The main contribution of this work is to produce a sampling-based global motion planner that generates different behaviors (in addition to locomotion) issued from environmental constraints. Several results of our algorithm in different environments are shown and its current limitations are discussed.