Data-driven biped control

Abstract

Figure 1: Our data-driven controller allows the physically-simulated biped character to reproduce challenging motor skills captured in motion data. AbstractWe present a dynamic controller to physically simulate underactuated three-dimensional full-body biped locomotion. Our datadriven controller takes motion capture reference data to reproduce realistic human locomotion through realtime physically based simulation. The key idea is modulating the reference trajectory continuously and seamlessly such that even a simple dynamic tracking controller can follow the reference trajectory while maintaining its balance. In our framework, biped control can be facilitated by a large array of existing data-driven animation techniques because our controller can take a stream of reference data generated on-thefly at runtime. We demonstrate the effectiveness of our approach through examples that allow bipeds to turn, spin, and walk while steering its direction interactively.

Human motion control with physically plausible foot contact models

“…The common criterion is the use of motion capture data, which enables the reproduction of plenty of details and nuances of original human motion. The QP-based optimization concept of a recent work [22] is similar to that of [18], (b) Lee et al [21], (c) Lee et al [22] ours, however, the purpose is different (i.e., simulating the human musculoskeletal system vs. handling dynamic footto-ground contact). The location, shape, and complexity of the velocity cone are also different, such that we can generate sharp foot turn motions that were not achievable using previous approaches [18,21,22].…”

“…Many optimization-based approaches [1,23,26,27,33] have designed to accurately address the dynamics of character, and compute joint torques and contact forces through minimizing the defined objective function while satisfying all physical constraints. As optimization-based approaches compute all joint torques at once, they often achieve more robustness when compared with proportional derivative (PD) servo-based walking controllers [10,13,18,21,31,34]. In addition, optimizationbased methods can handle higher levels of dynamic constraints to human motion to formulate physically complex problems such as controlling cartwheels and back-flips.…”

“…Tracking the reference motion capture data are advantageous for reproducing believable human motion [8,14,18,21,22,27,30,31,34]. Such data-driven approaches can generate high-quality and natural styles of characters because motion capture data contain abundant details and nuances of original human motion.…”

Adaptive locomotion on slopes and stairs using pelvic rotation

“…There are physics-based approaches that do not make use of motion capture data [6,7,27,38]. or that use only a few motion clips to generate a range of motions [16,19]. Various types of controllers have been developed using optimizations [5,[34][35][36]41].…”

Complex object manipulation with hierarchical optimal control

“…However, the approach that is currently quite promising is a hierarchical control approach [7][10], combined with model predictive control. Indeed, the most successful examples can be found in computer graphics where people are interested in creating interactive characters which function based upon physics rather than motion capture or kinematic data [18] [12][2] [6]. This approach has the potential to create very adaptable interactive characters which have unscripted and rich behaviors.…”