Composable Group Behaviors | Algorithms & Applications Group

Algorithms & Applications Group
Composable Group Behaviors

Composable Group Behaviors
supported by NSF
Jyh-Ming Lien, Samuel Rodriguez, Xinyu Tang, Nancy M. Amato
Project Alumni: Arnaud Masciotra

Creating animations with complex and realistic group behaviors can be a difficult and time consuming task. This generally involves associating all possible behaviors of agents and associating the behaviors with all environmental events.
In this work, we investigate methods to ease the process of producing realistic group behaviors. More specifically, the main goal of this research is to simulate group behaviors by automatically combining a given set of simple composable behaviors for applications such as games, virtual reality, robotics and biological/ecological simulation. The result of this research is an easy to use, adaptive and flexible framework for simulating group behaviors.

A system overview (click to enlarge the image)

Our system is built on the top of a roadmap based system.

A roadmap represents the connectivity of the feasible space in a given environment and is used as an internal representation of the world for agents.
Agents can modify the properties of the roadmap.
An agent can store location-specific information, such as the action that the agent takes, in a node of a roadmap.
The result of an action will be evaluated by a user defined utility function. The agent will select an action based on the past performance of the action.
The learned information can also be shared between agents when their relative states allow communication (e.g. proximity and other values).

As shown in the above figure, the proposed framework consists of several main components, i.e., agents, roadmaps, user defined behaviors and utility functions. We studied three types of behaviors:

a general pursuit/evasion behavior
a laser-tag game
a shepherding behavior

Pursuit/Evasion Behaviors

In this experiment, we show that an agent can adapt to its performance dramatically better by simply adding more behaviors to it.

The left figure shows the environment for this experiment. The prey are designed to only have the patrolling and evasion behaviors and normally walk back and forth along the patrolling path and start running away when predators are nearby. The predators can choose to either wait or search for prey in the whole environment if it does not see any. When they see a prey, they will start pursuing and then attacking it. Prey can run four times faster and see two times farther than the predator so that the prey becomes hard to catch.

Movie: Pursuit and Evasion (divx avi) 14.4 MB

Laser Tag

In the game of laser tag, agents work as teams in order to score points against agents from opposing teams.

An agent's goal is to shoot at opponents. After a specified number of hits against an agent, that agent can no longer participate in the game. Our laser tag environment, shown in the left figure, consists of two flock groups. The first flock group (called snipers) is equipped with a longer range of sight but a more restricted viewing angle. The viewing angle is the total angle in the heading direction that the flock member can see. The second flock group (called infantry) has a shorter range of sight but a much larger view angle.

Movie: Laser Tag (divx avi) 15.9 MB

Shepherding (see also Specialized Techniques for Shepherding Behaviors)

The idea of composable behaviors can be readily applied to compose shepherding behaviors from a set of simple primitive behaviors we call locomotions. The utility function of the shepherding behaviors is designed to encourage the shepherds to decrease the distance between the flock and the goal position and also to encourage the shepherds to keep the flock as one group. The figures below show a sequence of images captured from the herding simulation with a group of shepherd controls the motion of a group of flock.

Movie: Shepherding (divx avi) 9.3MB

Related Projects

Group Behaviors using Rule-Based Roadmaps
Shepherding Behaviors
Planning Among Moving Obstacles

Papers

A Framework for Planning Motion in Environments with Moving Obstacles, Sam Rodriguez, Jyh-Ming Lien, Nancy M. Amato, In Proc. IEEE Int. Conf. Intel. Rob. Syst. (IROS), pp. 3309-3314, Oct 2007. Also, Technical Report, TR06-010, Parasol Laboratory, Department of Computer Science, Texas A&M University, Sep 2007.
Proceedings(ps, pdf, abstract) Technical Report(ps, pdf, abstract)

Roadmap-Based Group Behaviors: Generation and Evaluation, Samuel Rodriguez, Robert Salazar, Troy McMahon, Nancy M. Amato, Technical Report, TR07-004, Parasol Laboratory, Department of Computer Science, Texas A&M University, Sep 2007.
Technical Report(ps, pdf, abstract)

Composable Group Behaviors, Jyh-Ming Lien, Samuel Rodriguez, Xinyu Tang, John Maffei, Arnaud Masciotra, Technical Report, TR05-006, Parasol Laboratory, Department of Computer Science, Texas A&M University, Sep 2005.
Technical Report(ps, pdf, abstract)

Shepherding Behaviors with Multiple Shepherds, Jyh-Ming Lien, Samuel Rodriguez, Jean-Philippe Malric, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), Apr 2005. Also, Technical Report, TR04-003, Parasol Laboratory, Department of Computer Science, Texas A&M University, Sep 2004.
Proceedings(ps, pdf, abstract) Technical Report(ps, pdf)

Swarming Behavior Using Probabilistic Roadmap Techniques, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, Lecture Notes in Computer Science, 3342/2005:112-125, Jan 2005.
Journal(ps, pdf, abstract)

Shepherding Behaviors, Jyh-Ming Lien, O. Burchan Bayazit, Ross T. Sowell, Samuel Rodriguez, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 4159-4164, New Orleans, Apr 2004. Also, Technical Report, TR03-006, Parasol Laboratory, Department of Computer Science, Texas A&M University, Nov 2003.
Proceedings(ps, pdf, abstract) Technical Report(ps, pdf)

Better Shepherding Behaviors Using Improved Shepherd Locomotion, Ross T. Sowell, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, Technical Report, TR03-009, Parasol Laboratory, Department of Computer Science, Texas A&M University, Aug 2003.
Technical Report(ps, pdf, abstract)

Better Group Behaviors in Complex Environments with Global Roadmaps, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, In Proc. Int. Conf. on the Sim. and Syn. of Living Sys. (Alife), pp. 362-370, Sydney, Australia, Dec 2002.
Proceedings(ps, pdf, abstract)

Better Group Behaviors using Rule-Based Roadmaps, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, In Proc. Int. Wkshp. on Alg. Found. of Rob. (WAFR), pp. 95-111, Nice, France, Dec 2002.
Proceedings(ps, pdf, abstract)

Roadmap-Based Flocking for Complex Environments, O. Burchan Bayazit, Jyh-Ming Lien, Nancy M. Amato, In Proc. Pacific Conf. on Computer Graphics and App. (PG), pp. 104-113, Beijing, China, Oct 2002. Also, Technical Report, TR02-003, Parasol Laboratory, Department of Computer Science, Texas A&M University, Apr 2002.
Proceedings(ps, pdf, abstract) Technical Report(ps, pdf, abstract)

Department of Computer Science and Engineering | Dwight Look College of Engineering | Texas A&M University
Privacy statement: Computer Science and Engineering Engineering TAMU