Search Results

The simulation of natural human motion remains a challenge. Considerable effort has been spent searching for methods that accurately predict how humans move under different task and environmental conditions. Motivation for these investigations comes from both an interest to understand how the human biological system controls complex motions, and by a need for reasonable motion prediction for a wide range of animation needs. A list of performance criteria is used to review the more popular control methods that have been described in the literature. Apparent from the review is that currently no single method exists for general human movement simulation, yet the eventual construct of a successful controller is becoming more apparent.

A common request from users of human simulation tools is for functionality that allows rapid definition of task sequences and subsequent, largely automatic, “what if” scenario evaluation. A review of the high level requirements collected from feedback on our previous implementation forays of such systems is presented, and a new framework is suggested to address the broad set of identified requirements. The framework, called the Task Simulation Builder, is characterized by a flexible and extensible architecture centered around a learning type task controller. Minimum user input is required to describe a scenario of events and obtain a simulation output, yet through output refinement feedback, the resulting simulation can progressively be made as detailed as needed. The overall goal of the system is to improve the efficiency of human modeling use in production environments.