BehaveRT: A GPU-Based Library for Autonomous Characters

Abstract

In this work, we present a GPU-based library, called BehaveRT, for the definition, real-time simulation, and visualization of large communities of individuals. We implemented a modular flexible and extensible architecture based on a plug-in infrastructure that enables the creation of a behavior engine system core. We used Compute Unified Device Architecture to perform parallel programming and specific memory optimization techniques to exploit the computational power of commodity graphics hardware, enabling developers to focus on the design and implementation of behavioral models. This paper illustrates the architecture of BehaveRT, the core plug-ins, and some case studies. In particular, we show two high level behavioral models, picture and shape flocking, that generate images and shapes in 3D space by coordinating the positions and color-coding of individuals. We, then, present an environment discretization case study of the interaction of a community with generic virtual scenes such as irregular terrains and buildings

Multi-agent simulation on multiple GPUs

“…Alg 1 column is the time to complete ''setup field'' kernel in Section 3.3.1. In addition, as a comparison with related works, the Radix Sort column contains the execution time of the sorting kernel of [41] used as one of the four steps for achieving the same result as our Algorithm 1. The table shows the result of our strategy in increasing the computational complexity of each model.…”

“…Besides, the implementation of Smoothed Particle Hydrodynamics (SPH) on a single GPU [39] also uses the spartial data structure for neighbor search in a bounded space. [40] also attempted to provide a reusable set of kernels for finding neighbors in the BehaveRT library [41]. The approach introduces 4 kernels, namely hashing, sorting, data reordering and cells counting.…”

ADAPT: The Agent Developmentand Prototyping Testbed

“…The problem of managing a character's behavior can be represented with decision networks [41], cognitive models [42], goal-oriented action planning [43], [44], or via learning [45]. Very simple agents can also be simulated on a massive scale using GPU processing [46]. Recent work [47], [48], [49] proposes an event-centric authoring paradigm to facilitate multi-actor interactions with contextual awareness based on agent type and event location.…”

Software entities as bird flocks and fish schools

“…Another advantage is that the number of clusters to recovery is not a priori specified. Finally, our distributed behavioral model can be easily implemented on GPU (Graphical Processor Unit) [21] to improve its scalability and to reduce energy consumption (e.g., [18]). …”