Home research People General Info Seminars Resources Intranet
| Algorithms & Applcations Group | Home | Research | Publications | People | Resources | News
Algorithms & Applications Group
Evacuation Planning

Project Personnel:Samuel Rodriguez,Nancy Amato

In this work, we use our framework for simulating and controlling communities of characters that can interact with each other and their environment, and can dynamically react to changes. An evacuation situation, is an example where complex interaction is needed between many agents in a scene. In an evacuation scenario, agents attempt to flee an area while some agents (for example law enforcement) may attempt to direct the agents to safe areas. The direction given can either serve as a guide to assist the evacuating agents or as a command (a route the agents should not deviate from). Similar situations can be seen when traffic lights are out and law enforcement has to direct traffic.

We are interested in studying is the evacuation of agents under a variety of conditions. By being able to simulate agents that are evacuating an area, we can study the effects of things such as exit placements, available exits, and agents directing the evacuating agents through the environment. In these scenarios we have some agents that are attempting to evacuate the first floor of a building. The agents have to find paths to the safe areas. We start by studying a building evacuation consisting of several hundred evacuating agents that take into account barriers and directing agents placed throughout the environment to control the evacuating agents.

Environmental Effect on Egress Simulation

Evacuation and egress simulations can be a useful tool in determining the effect of design decisions on the flow of agent movement. This type of simulation can be used to determine beforehand the effect of such design decisions and potential improvements that can be made. In this work, we look at how agent egress is effected by the environment in real world and large scale virtual environments and investigate metrics to analyze the flow. Our work differs from many evacuation systems in that we handle real world environments, support grouping restrictions betweens agents (e.g., families or other social groups traveling together), and model the scenario of agents moving with multiple modes of transportation with physically realistic dynamics (e.g., individuals walk from a building to their own cars and leave only when all people in the group arrive). We also focus on the motion strategies necessary to allow agents to navigate in these complex environments.

Scenario: 400 agents in a multi-level building undergoing egress experiencing different architectural components and design factors.

  • Varying door types (mp4)
  • Varying pillar placement (mp4)
  • Scenario: Pedestrians and vehicles performing egress behavior. Pedestrians can be associated with vehicles which have to wait for the associated pedestrians to arrive.

  • 200 pedestrians, 40 vehicles, lot configuration 1 (mp4)
  • 200 pedestrians, 40 vehicles, lot configuration 2 (mp4)
  • Example with pedestrians in Langford A/B and nearby actual parking lot (mp4)

  • Evacuation in Complex Scenarios

    We investigate utilization of roadmaps in a general evacuation planning system for complex 3D environments. The problem consists of heterogeneous groups of agents whose behaviors and properties affect usage of the environment when creating evacuation plans. This planning system includes behaviors for those agents evacuating and directors that may be guiding the agents to improve evacuation. One aspect we focus on is modeling different forms of direction and communication between agents.

    Scenario: 500 agents in a basic multi-level environment evacuating under different conditions

  • Homogeneous agents
  • Heterogeneous agents utilizing a shared roadmap
  • Heterogeneous agents given direction
  • Animation (full 30MB) (small 7MB)
  • Scenario: 500 agents in a complex 3-level building evacuating under different conditions

  • Shared Roadmap
  • Direction Away from Dangerous Area
  • Direction Away from Corner Exit
  • Animation (full 80MB) (small 12MB)

  • To show how our system can extend to much more complex scenarios we show an example evacuation of a stadium.

    Scenario: 20,000 agents in a complex stadium-like environment evacuating to safe locations.

  • Shared Roadmap
  • Large-scale evacuation
  • Agents with heterogeneous parameters
  • Animation (full 40MB)

  • Another large-scale evacuation, this one of the Bright and Langford buildings which house our Computer Science and Engineering Department and Architecture Department on the Texas A&M University campus.

    Scenario: 1,100 agents in a complex set of buildings evacuating to safe locations.

  • Large-scale evacuation of actual buildings
  • Shared Roadmap
  • Agents with heterogeneous parameters
  • Animation mp4 (12MB) avi (109MB)

  • Evacuation Example

    This is an example environment which shows many of the capabilities that we are able to simulate, some unique to our system. This environment consists of two rooms with a number of area types available, depending on the scenario. The area types available range from three exits in the main room, two safe areas (at the lower and right portions of the environment), and a potential dangerous area in the second room. Thirty evacuating agents begin the simulation clustered at the center of the lower room. In scenarios (b)-(e) a director is present to guide evacuation. A full evacuation is shown in scenario (a).
    (a) Full Evacuation.
    (b) With Director (2EXs).
    (c) With Director (2EXs+1SA).
    (d) Evacuation w/ Director (Grouping+2EXs+1SA).
    (e) Evacuation w/ Director (2EXs+1SA+1DA).

    Evacuation Bright

    Here we show evacuation results for agents evacuating a building under different evacuation conditions. The test environment is the first floor of a building at Texas A&M University housing our department. Agents that will be evacuating are randomly placed in rooms throughout the building. As the simulation progresses, the agents try to evacuate the building, utilizing the roadmap to find paths to safe areas. We are able to look at many different evacuation scenarios and analyze the results. We show experimental results with 500 agents evacuating the first floor, varying the number of available exits, the amount and type of direction given during the evacuation and which exits, if any, are being prohibited or blocked.


    Example scenario 1 - Evacuation

    In the evacuation shown here, the agents attempt to evacuate by finding the nearest available exit. The agents try to find a path to this nearest safe exit and then to the safe location.
    Example of agents evacuating the Bright building (5 exits).
    Example of agents evacuating the Bright building (4 exits).
    Example of agents evacuating the Bright building (3 exits).


    Example scenario 2 - Evacuation with Direction

    In this scenario we have agents that are directing the evacuating agents away from one side of the environment. As the agents come within range of the directing agent, they are told of a potentially dangerous area and prevented from exiting from one side of the building. Agents that are directed away from an area then have to replan to find a route to a safe area. In the first animation, 1 directing agent is present to direct agents away from one available exit. In the second animation, 4 directing agents are placed throughout the environment and tell the evacuating agents about the area to avoid. Using this kind of simulating we can study the flow of agents and potential problems that could arise during evacuation. This could also give insight into how many agents should be present to direct and the placement of directing agents to enable a good evacuation procedure the evacuating agents.
    Example of agents evacuating with one barrier.
    Example of agents evacuating with two barriers.
    Example of agents evacuating with two global directors.
    Example of agents evacuating with four global directors.
    Example of agents evacuating with five global directors.
    Example of agents evacuating with one global director and two points of direction.
    Example of agents evacuating with one global director and four points of direction.

    Regrouping Example

    We show this regrouping example to show the versatility of our framework. In this scenario agents are dispersed around an environment and regroup to safe locations defined throughout the environment. The animations show four different scenarios we were able to generate evacuation plans for. In the first scenario, the agents regroup to the nearest safe area. In the second example, directing agents are placed at exits E3 and E5 to simulate a partial blockage of the corridor. The result is that the agents still use both safe areas although they do not pass through the blocked corridor. In the third and fourth scenario, two advanced directing agents are placed at the locations shown in the figure near E3 and E5. These agents, with a larger view radius, are alerting the evacuating agents to the dangerous areas present in the corridor. The result is that the agents regroup at the safe areas near the upper corridor in the environment. In the fourth scenario grouping is used as a restriction for the agents.
    (a) Basic Regroup.
    (a) Partial blockage of Lower Corridor.
    (a) Full Block of Lower Corridor.
    (a) Full Block of Lower Corridor (Grouping).

    Directional Exp1: Passages

    The first environment in our directional study consists of many passages leading to four exits at the top of the environment. The agents must travel from their initial locations, at the lower portions of the environment, to a safe area at the top of environment.
    Scenarios:

    Directional Exp2: Museum

    In the museum environment, evacuating agents are initially placed at random locations through the environment with a bias toward the left portion of the environment. In most of these examples, a director prevents passage through the left exit.
    Scenarios:

      A full evacuation where agents have 4 known available exits (N,S,E,W). [Scenario (a)]
    • Same as in (a), but a director at Exit E acting as barrier with: [Scenario (b) P_d=1.0] [Scenario (c) P_d=0.8]
    • Same as in (b--c) but director at Exit E is coordinated assigning to three other exits (N,S,W) (d) P_d=1.0, (e) P_d=0.8: [Scenario (d) P_d=1.0] [Scenario (e) P_d=0.8]
    • Similar to (b) however with 4 local coordinated directors also present at NE, NW, SW, and SE exits to coordinate evacuation to nearest two exits. (f) P_d=1.0, (g) P_d=0.8: [Scenario (f) P_d=1.0] [Scenario (g) P_d=0.8]
    • 1 barrier + 3 coordinated directors placed for maximum influence with P_d=1.0: Director at center of environment biasing toward right side of environment (Exit 3,6,7); Director at 5 biasing toward NE exit; Director at 8 biasing toward SE exit, [Scenario (h)]

    Reciprocally-Rotating Velocity Obstacles, Andrew Giese, Daniel Latypov, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. to appear, Hong Kong, China, Jun 2014. Also, Technical Report, TR13-009, Parasol Laboratory, Department of Computer Science, Texas A&M University, College Station, TX U.S.A., Oct 2013.
    Proceedings(ps, pdf, abstract) Technical Report(ps, pdf, abstract)

    Reciprocally-Rotating Velocity Obstacles, Andrew Giese, Masters Thesis, Department of Computer Science and Engineering, Texas A&M University, College Station, USA, May 2014.
    Masters Thesis(ps, pdf, abstract)

    Multi-Robot Caravanning, Jory Denny, Andrew Giese, Aditya Mahadevan, Arnaud Marfaing, Rachel Glockenmeier, Colton Revia, Samuel Rodriguez, Nancy M. Amato, In Proc. IEEE Int. Conf. Intel. Rob. Syst. (IROS), pp. 5722 - 5729, Tokyo, Japan, Nov 2013.
    Proceedings(pdf, abstract)

    Optimizing Aspects of Pedestrian Traffic in Building Designs, Samuel Rodriguez, Yinghua Zhang, Nicholas Gans, Nancy M. Amato, In Proc. IEEE Int. Conf. Intel. Rob. Syst. (IROS), Nov 2013.
    Proceedings(pdf, abstract)

    Improving Aggregate Behavior in Parking Lots with Appropriate Local Maneuvers, Samuel Rodriguez, Andrew Giese, Nancy M. Amato, In Proc. IEEE Int. Conf. Intel. Rob. Syst. (IROS), Nov 2013.
    Proceedings(pdf, abstract)

    Environmental Effect on Egress Simulation, Samuel Rodriguez, Andrew Giese, Nancy M. Amato, Saeid Zarrinmehr, Firas Al-Douri, Mark Clayton, In Proc. of the 5th Intern. Conf. on Motion in Games (MIG), 2012, in Lecture Notes in Computer Science (LNCS), pp. to appear, Rennes, Brittany, France, Nov 2012.
    Proceedings(ps, pdf, abstract)

    A Sampling-Based Approach to Probabilistic Pursuit Evasion, Aditya Mahadevan, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 3192 - 3199, St. Paul, Minnesota, USA, May 2012.
    Proceedings(pdf, abstract)

    Roadmap-Based Techniques for Modeling Group Behaviors in Multi-Agent Systems, Samuel Rodriguez, Ph.D. Thesis, Department of Computer Science and Engineering, Texas A&M University, Jan 2012.
    Ph.D. Thesis(ps, pdf, abstract)

    Roadmap-Based Level Clearing of Buildings, Samuel Rodriguez, Nancy M. Amato, In Proc. of the 4th Intern. Conf. on Motion in Games (MIG), 2011, in Lecture Notes in Computer Science (LNCS), pp. 340-352, Edinburgh, UK, Oct 2011.
    Proceedings(ps, pdf, abstract)

    Toward Realistic Pursuit-Evasion Using a Roadmap-Based Approach, Samuel Rodriguez, Jory Denny, Juan Burgos, Aditya Mahadevan, Kasra Manavi, Luke Murray, Anton Kodochygov, Takis Zourntos, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 1738-1745, May 2011.
    Proceedings(ps, pdf, abstract)

    Roadmap-Based Pursuit-Evasion in 3D Structures, Samuel Rodriguez, Jory Denny, Aditya Mahadevan, Jeremy (Cong-Trung) Vu, Juan Burgos, Takis Zourntos, Nancy M. Amato, In Proc. of 24th Intern. Conf. on Computer Animation and Social Agents (CASA), 2011, in Transactions on Edutainment, pp. to appear, May 2011.
    Proceedings(ps, pdf, abstract)

    Utilizing Roadmaps in Evacuation Planning, Samuel Rodriguez, Nancy M. Amato, In Proc. of 24th Intern. Conf. on Computer Animation and Social Agents (CASA), 2011, in Intern. J. of Virtual Reality (IJVR), pp. 67-73, May 2011.
    Proceedings(ps, pdf, abstract)

    Toward Simulating Realistic Pursuit-Evasion Using a Roadmap-Based Approach, Samuel Rodriguez, Jory Denny, Takis Zourntos, Nancy M. Amato, In Proc. of the 3rd Intern. Conf. on Motion in Games (MIG), 2010, in Lecture Notes in Computer Science (LNCS), pp. 82-93, Nov 2010.
    Proceedings(ps, pdf, abstract)

    Behavior-Based Evacuation Planning, Sam Rodriguez, Nancy M. Amato, In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 350-355, Anchorage, AK, May 2010.
    Proceedings(ps, pdf, abstract)

    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.
    Proceedings(ps, pdf, abstract)

    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, 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)