Reynolds的自主角色导航与生活化行为设计

需积分: 9 92 浏览量更新于2024-07-22 收藏 199KB PDF 举报

本文探讨了动画与游戏中的自主角色（Autonomous Characters）所面临的动态需求：即如何以类似生物的方式在虚拟环境中进行自然、即兴的导航。作者 Craig W. Reynolds 是来自 Sony Computer Entertainment America 的专家，他的研究聚焦于“Steering Behaviors”，这些行为策略独立于角色的具体移动方式，如步行、飞行或驾驶。 Steering Behaviors 是一系列设计用于模拟真实世界中生物体（如鸟类或车辆）在面对障碍、目标追求、群体行为和环境交互时的决策和动作。它们主要包括路径规划、路径跟踪、追逐、躲避、障碍物和碰撞避免等技术。这些行为能够组合起来，帮助角色实现更高层次的目标，比如从一个位置到达另一个位置的同时避开障碍，沿着指定的走廊移动，或者加入特定的群体。文章将角色的运动行为划分为三个层次： 1. **基础层面（Basic Level）**：处理角色的基本移动和转向，确保其在物理空间中的定位和基本控制，如直线前进、停止、转弯等。 2. **反应层面（ Reactive Level）**：基于即时反馈进行调整，例如看到障碍物时改变方向或速度，遇到群体时进行互动。 3. **智能层面（Intelligent Level）**：包含更高级的策略，如路径选择、群体协同和情境适应，使角色能够在复杂环境中做出灵活而符合逻辑的行为，如模拟社会性行为（如领队或跟随）或集体行动（如群集和疏散）。 Reynolds 的工作强调了角色的“situated”和“embodied”特性，意味着它们的行为不仅依赖于预设的编程，还取决于它们在虚拟世界中的具体位置、感知能力和环境状态。此外，他还讨论了人工智能（Artificial Life）在其中的应用，以及如何通过模仿现实世界的生物群体行为（如 flocking 或 schooling）来增强游戏的真实感和沉浸性。这篇论文的关键字包括 Animation Techniques（动画技术）、Virtual/Interactive Environments（虚拟/互动环境）、Games（游戏）、Simulation（模拟）、Behavioral Animation（行为动画）、Autonomous Agent（自主代理）、Improvvisation（即兴表演）等，它为设计更加逼真且富有表现力的自主角色提供了理论基础和实用方法，对游戏开发者和虚拟世界构建者具有重要的指导意义。

action selection and with Tinsley Galyean in [Blumberg 95] discussed the design for a VR

character capable of both autonomous improvisation and response to external direction. One

application of these characters was in the ALIVE system [Maes 95] by Patties Maes et al. The

Improv system by Ken Perlin and Athomas Goldberg [Perlin 96] also covers the gamut from

locomotion to action selection, but uses a unique approach based on behavioral scripting and

Perlin’s 1985 procedural synthesis of textures [Perlin 85] applied to motion. James Cremer

and colleagues have created autonomous drivers to serve as “extras” creating ambient traffic

in interactive automobile driving simulators [Cremer 96]. Robin Green (of Bullfrog/EA) has

developed a mature system for autonomous characters used in Dungeon Keeper 2 which was

inspired in part by an early draft of this paper. Dave Pottinger has provides a detailed

discussion of steering and coordination for groups of characters in games [Pottinger 1999].

Locomotion

Locomotion is the bottom of the three level behavioral hierarchy described above. The

locomotion layer represents a character’s embodiment. It converts control signals from the

steering layer into motion of the character’s “body.” This motion is subject to constraints

imposed by the body’s physically-based model, such as the interaction of momentum and

strength (limitation of forces that can be applied by the body).

As described above, a cowboy’s horse can be considered as an example of the locomotion

layer. The rider’s steering decisions are conveyed via simple control signals to the horse who

converts them into motion. The point of making the abstract distinction between steering and

locomotion is to anticipate “plugging in” a new locomotion module. Imagine lifting the rider off

of the horse and placing him on a cross-country motorcycle. The goal selection and steering

behavior remain the same. All that has changed is the mechanism for mapping the control

signals (go faster, turn right, ...) into motion. Originally it involved legged locomotion (balance,

bones, muscles) and now it involves wheeled locomotion (engine, wheels, brakes). The role of

the rider is unchanged.

This suggests that with an appropriate convention for communicating control signals, steering

behaviors can be completely independent of the specific locomotion scheme. Although in

practice it is necessary to compensate for the “agility” and different “handing characteristics” of

individual locomotion systems. This can be done by adjusting tuning parameters for a given

locomotion scheme (which is the approach taken in the steering behaviors described below) or

by using an adaptive, self-calibrating technique (the way a human driver quickly adapts to the

characteristics of an unfamiliar automobile). In the first case a steering behavior might

determine via its a priori tuning that the character’s speed in a given situation should be 23

mph, in the second case it might say “slow down a bit” until the same result was obtained.

The locomotion of an autonomous character can be based on, or independent from, its

animated portrayal. A character could be represented by a physically-based dynamically

balanced simulation of walking, providing both realistic animation and behavioral locomotion.

Or a character may have a very simple locomotion model (like described in the next section) to

which a static (say a spaceship) or pre-animated (like a human figure performing a walk cycle)

portrayal is attached. A hybrid approach is to use a simple locomotion model and an adaptive

animation model, like an inverse-kinematics driven walk cycle, to bridge the gap between

abstract locomotion and concrete terrain. Finally, locomotion can be restricted to the motion

剩余20页未读，继续阅读

cugfjp

粉丝: 2
资源: 1

Reynolds的自主角色导航与生活化行为设计

steering engine

第一部分AI人工智能vc6项目源码

3D-steering-behaviors:使用 webgl 演示转向行为

Create-a-Hockey-Game-Using-Steering-Behaviors:使用转向行为创建曲棍球比赛

steering system

Steering-Behaviours.zip_The Seek_inertia _steering_steering beha

Steering-Behaviours

Steering-wheel

steering_functions

eps.rar_eps steering_steering_汽车_灵敏度matlab

最新资源