Shawn Singh

Egocentric affordance fields in pedestrian steering

In this paper we propose a general framework for local path-planning and steering that can be easily extended to perform high-level behaviors. Our framework is based on the concept of affordances - the possible ways an agent can interact with its environment. Each agent perceives the environment through a set of vector and scalar fields that are represented in the agents local space. This egocentric property allows us to efficiently compute a local space-time plan. We then use these perception fields to compute a fitness measure for every possible action, known as an affordance field. The action that has the optimal value in the affordance field is the agents steering decision. Using our framework, we demonstrate autonomous virtual pedestrians that perform steering and path planning in unknown environments along with the emergence of high-level responses to never seen before situations.

A Behavior-Authoring Framework for Multiactor Simulations

Interest has been growing in the behavioral animation of autonomous actors in virtual worlds. However, authoring complicated interactions between multiple actors in a way that balances control flexibility and automation remains a considerable challenge. A proposed behavior-authoring framework gives users complete control over the domain of the system: the state space, action space, and cost of executing actions. To specialize actors, the framework uses effect and cost modifiers, which modify existing action definitions, and constraints, which prune action choices in a state-dependent manner. The framework groups actors with common or conflicting goals to form a composite domain, and a multiagent planner generates complicated interactions between multiple actors. The Web extra is a video that shows how multiactor simulations should aim to strike a happy medium between the automation of generation and the flexibility of specification.

Scenario space: characterizing coverage, quality, and failure of steering algorithms

Navigation and steering in complex dynamically changing environments is a challenging research problem, and a fundamental aspect of immersive virtual worlds. While there exist a wide variety of approaches for navigation and steering, there is no definitive solution for evaluating and analyzing steering algorithms. Evaluating a steering algorithm involves two major challenges: (a) characterizing and generating the space of possible scenarios that the algorithm must solve, and (b) defining evaluation criteria (metrics) and applying them to the solution. In this paper, we address both of these challenges. First, we characterize and analyze the complete space of steering scenarios that an agent may encounter in dynamic situations. Then, we propose the representative scenario space and a sampling method that can generate subsets of the representative space with good statistical properties. We also propose a new set of metrics and a statistically robust approach to determining the coverage and the quality of a steering algorithm in this space. We demonstrate the effectiveness of our approach on three state of the art techniques. Our results show that these methods can only solve 60% of the scenarios in the representative scenario space.

Situation agents: agent-based externalized steering logic

Realistic character animation requires elaborate rigging built on top of high quality 3D models. Sophisticated anatomically based rigs are often the choice of visual effect studios where life-like animation of CG characters is the primary objective. However, rigging a character with a muscular-skeletal system is very involving and time-consuming process, even for professionals. Although, there have been recent research efforts to automate either all or some parts of the rigging process, the complexity of anatomically based rigging nonetheless opens up new research challenges. We propose a new method to automate anatomically based rigging that transfers an existing rig of one character to another. The method is based on a data interpolation in the surface and volume domain, where various rigging elements can be transferred between different models. As it only requires a small number of corresponding input feature points, users can produce highly detailed rigs for a variety of desired character with ease. Copyright

An Open Framework for Developing, Evaluating, and Sharing Steering Algorithms

There are very few software frameworks for steering behaviors that are publicly available for developing, evaluating, and sharing steering algorithms. Furthermore, there is no widely accepted methodology for how to evaluate results of agent steering simulations. This situation makes it difficult to identify the real underlying challenges in agent simulations and future research directions to advance the state of the art. With the hope of encouraging community participation to address these issues, we have released SteerSuite , a flexible but easy-to-use set of tools, libraries, and test cases for steering behaviors. The software includes enhanced test cases, an improved version of SteerBench, a modular simulation engine, a novel steering algorithm, and more. Care has been taken to make SteerSuite practical and easy-to-use, yet flexible and forward-looking, to challenge researchers and developers to advance the state of the art in steering.

Parallelized egocentric fields for autonomous navigation

In this paper, we propose a general framework for local path-planning and steering that can be easily extended to perform high-level behaviors. Our framework is based on the concept of affordances: the possible ways an agent can interact with its environment. Each agent perceives the environment through a set of vector and scalar fields that are represented in the agent’s local space. This egocentric property allows us to efficiently compute a local space-time plan and has better parallel scalability than a global fields approach. We then use these perception fields to compute a fitness measure for every possible action, defined as an affordance field. The action that has the optimal value in the affordance field is the agent’s steering decision. We propose an extension to a linear space-time prediction model for dynamic collision avoidance and present our parallelization results on multicore systems. We analyze and evaluate our framework using a comprehensive suite of test cases provided in SteerBench and demonstrate autonomous virtual pedestrians that perform steering and path planning in unknown environments along with the emergence of high-level responses to never seen before situations.

SteerBug: an interactive framework for specifying and detecting steering behaviors

The size of crowds that modern computer games and urban simulations are capable of handling has given rise to the challenging problem of debugging and testing massive simulations of autonomous agents. In this paper, we propose SteerBug: an interactive framework for specifying and detecting steering behaviors. Our framework computes a set of time-varying metrics for agents and their environment, which characterize steering behaviors. We identify behaviors of interest by applying conditions (rules) or user defined sketches on the associated metrics. The behaviors we can specify and detect include unnatural steering, plainly incorrect results, or application-specific behaviors of interest. Our framework is extensible and independent of the specifics of any steering approach. To our knowledge, this is the first work that aims to provide a computational framework for specifying and detecting crowd behaviors in animation.

Watch Out! A Framework for Evaluating Steering Behaviors

Interactive virtual worlds feature dynamic characters that must navigate through a variety of landscapes populated with various obstacles and other agents. The process of navigating to a desired lo- cation within a dynamic environment is the problem of steering . While there are many approaches to steering, to our knowledge there is no standard way of evaluating and comparing the quality of such solutions. To address this, we propose a diverse set of benchmarks and a flexi- ble method of evaluation that can be used to compare different steering algorithms. We discuss the challenges and criteria for objectively eval- uating steering behaviors and describe the metrics and scoring method used in our benchmark evaluation. We hope that, with constructive feed- back from the community, our framework will eventually evolve into a standard and comprehensive approach to debug, compare and provide an overall assessment of the effectiveness of steering algorithms.

Footstep navigation for dynamic crowds

The majority of previous crowd steering algorithms model each character as an oriented particle that moves by choosing a force or velocity vector. In many cases, orientation is heuristically chosen to be the same as the particles velocity. This approach has the two key disadvantages: Limited locomotion constraints: Vector commands do not account for constraints of real human movement. Trajectories may have discontinuous velocities, oscillations, awkward orientations, or may try to move a character unnaturally, and these side-effects make it harder to animate the character intelligently. or example, a character moving forward cannot easily step to the right when its left foot is in the air (swing phase). Limited navigation control: It is common to assume that an animation system will automatically know how to interpret a vector-based steering decision. However a vector does not have enough information to indicate appropriate subtle maneuvers, such as side-stepping versus reorienting the torso, stepping backwards versus turning around, stopping and starting, planting a foot to change momentum quickly, or carefully placing footsteps in exact locations. These details are critical to depicting a characters local steering intelligence, and thus it is appropriate for steering to have better control.

SIMD Packet Techniques for Photon Mapping

We present a novel photon mapping framework that uses single instruction, multiple data (SIMD) parallelism to accelerate the final gathering phase of photon mapping. By using SIMD instructions, four coherent tasks can be computed in parallel using almost the same memory traffic as it would cost to process one task alone. This approach has been very successful for real-time ray tracing, but until now it has been unclear how to effectively apply the same approach to final gathering. Our solution is to use sample-point density estimation instead of k-nearest neighbor density estimation, a technique drawn from reverse photon mapping. Sample-point estimation removes the overheads that make SIMD instructions impractical, while retaining the same benefits and image quality as traditional photon mapping. Additionally, an important question arises whether it is better to use forward or reverse photon mapping. In an interactive context, classical asymptotic algorithmic analysis is not enough to compare the two algorithms. We provide a novel asymptotic bandwidth analysis, which addresses more issues found in practice. The analysis motivates the use of forward photon mapping when using SIMD parallelism as well as partial reordering for improved scalability. The resulting framework can achieve interactive rates for photon mapping at low resolutions, including the time it takes to trace photons and build the photon map.