Posture Prediction and Optimization - Human Motion Simulation: Predictive Dynamics - page 65

objective function because the DOT does not handle the constrained optimization
well. In fact, the optimization yields similar results with this target point when
the objective terms are omitted entirely. This suggests that the distance term in
Equation (3.24) dominates the objective. However, if the multiplier (1000) is
reduced to rectify this problem, then the constraint is not adequately satisfied.
Conceptually and practically, the objective needs to be treated independently.
Furthermore, using a larger-DOF model that allows more realistic movement in
the upper torso and clavicle will improve the results.
3.14 Multi-objective problem statement
In this section, we formulate the posture prediction optimization problem using
the above-described model. MOO is used first to develop new human perfor-
mance measures. It is then used to combine these measures that serve as multiple
objective functions in the final optimization formulation.
3.15 Design variables and constraints
As suggested earlier, the design variables for the final MOO problem are q i ,
which indicate joint angles in units of degrees. The vector q represents the conse-
quent posture. Because listing values for all of the joint angles with each pre-
dicted posture can be cumbersome and unrevealing, and therefore, results in the
design space are depicted with actual pictures of the avatar.
The first constraint, called the distance constraint, requires the end-effector to
contact the target point. There is one distance constraint for each end-effector. In
addition, each generalized coordinate is constrained to lie within predetermined lim-
its; q i represents the upper limit for q i , and q i represents the lower limit. These lim-
its ensure that the virtual human does not assume an unrealistic posture given the
nature of actual human joints. Finally, we will demonstrate that the orientation of a
body part can be controlled by implementing an additional orientation constraint.
3.16 Concluding remarks
This chapter has introduced the concept of posture prediction as an effective and
robust method that yields natural human motion.
The main concept from this chapter is the use of optimization to solve human
motion problems. We have shown how optimization lends itself well to answering
the questions: “What would a human do under these constraints?” and “What is
the best solution given those constraints?”
Posture prediction using optimization-based methods is effective in addressing
the prediction of human motion and yields the following results: