Environment-aware Motion Matching

TL;DR

Environment-aware Motion Matching integrates an environment-driven, two-stage search with a simple collision proxy to produce real-time, full-body animation that naturally adapts pose and trajectory in crowded and obstacle-rich scenes. By separating query features (pose/trajectory) from environment features (collision penalties) and using a log-barrier penalization, the method achieves robust obstacle avoidance while maintaining motion quality. Key contributions include a 2D ellipse-based body representation, a two-tier feature search with temporal coherence optimizations, and seamless integration into crowd simulations using a single actor mocap database. The approach yields lightweight memory requirements, no training, and straightforward adaptability to new animation styles, making it well-suited for games, VR/AR, and large-scale crowds.

Abstract

Interactive applications demand believable characters that respond naturally to dynamic environments. Traditional character animation techniques often struggle to handle arbitrary situations, leading to a growing trend of dynamically selecting motion-captured animations based on predefined features. While Motion Matching has proven effective for locomotion by aligning to target trajectories, animating environment interactions and crowd behaviors remains challenging due to the need to consider surrounding elements. Existing approaches often involve manual setup or lack the naturalism of motion capture. Furthermore, in crowd animation, body animation is frequently treated as a separate process from trajectory planning, leading to inconsistencies between body pose and root motion. To address these limitations, we present Environment-aware Motion Matching, a novel real-time system for full-body character animation that dynamically adapts to obstacles and other agents, emphasizing the bidirectional relationship between pose and trajectory. In a preprocessing step, we extract shape, pose, and trajectory features from a motion capture database. At runtime, we perform an efficient search that matches user input and current pose while penalizing collisions with a dynamic environment. Our method allows characters to naturally adjust their pose and trajectory to navigate crowded scenes.

Figures (21)

• Figure 1: Overview of our Environment-aware Motion Matching pipeline. The system operates in two distinct stages: a preprocessing phase (Section \ref{['sec:method:preprocess']}) and a real-time controller (Section \ref{['sec:method:controller']}). The real-time controller utilizes user input and the current character pose to construct a query vector, which is then compared against the query features. Simultaneously, environment features guide the search by computing dynamic obstacle penalizations.
• Figure 2: The log-barrier function used for obstacle penalization. This plot illustrates the penalty $f(d) = -(t - d)^p \log \left( \frac{d}{t} \right)$ as a function of the distance $d$ to an obstacle. The penalty is zero when $d \ge t$ (the obstacle distance threshold), and increases exponentially as $d$ approaches zero. The parameter $p$ controls the steepness of this exponential growth, ensuring a strong repulsion as the character gets close to an obstacle.
• Figure 3: Purple arrows indicate the target trajectory derived from user input. Our method, while being aware of the cone obstacle, dynamically adjusts the character's path, causing it to step right to avoid collision. This demonstrates how the system integrates environmental constraints into root motion, ultimately enabling the character to reach the target trajectory while naturally avoiding obstacles.
• Figure 4: Comparison of our Environment-aware Motion Matching (top row) with standard Motion Matching (bottom row) in an obstacle-filled scene. The black line indicates the target path, and purple points denote future target positions. Blue points illustrate the system's selected future character positions. Our method (top) dynamically detours to avoid the central obstacle, showcasing its environment awareness and the coupled pose-trajectory selection. In contrast, standard Motion Matching (bottom) disregards the obstacle, leading to collisions. This highlights our system's bidirectional control, in contrast to typical decoupled animation pipelines.
• Figure 5: Character navigating a high-density crowd. The ligth blue t-shirt character walks slowly and carefully between other stationary characters, demonstrating the system's ability to find and adapt to paths within extremely confined, high-density environments.