Multi-agent based path planning for first responders among moving obstacles

https://doi.org/10.1016/j.compenvurbsys.2015.11.001Get rights and content

Highlights

  • We investigate four navigation cases for first responders among moving obstacles.

  • Develop a multi‐agent system to support path finding among moving obstacles

  • Develop a data model to manage data for multiple responders with multiple tasks

  • Provide a set of algorithms for different types of path panning problems

  • Define a multi‐objective function to minimize the delays caused by moving obstacles

Abstract

Natural or man-made disasters can cause different kinds of moving obstacles (e.g., fires, plumes, floods), which make some parts of the road network temporarily unavailable. After such incidents occur, responders have to go to different destinations to perform their tasks in the environment affected by the disaster. Therefore they need a path planner that is capable of dealing with such moving obstacles, as well as generating and coordinating their routes quickly and efficiently. In this paper, we present a novel approach for using a multi-agent system for navigating one or multiple responders to one or multiple destinations in the presence of moving obstacles. Our navigation system supports information collection from hazard simulations, spatio-temporal data processing and analysis, connection with a geo-database, and route generation in dynamic environments affected by disasters. We design and develop a set of software geospatial agents that assist emergency actors in dealing with the spatio-temporal data required for emergency navigation, based on their roles in the disaster response. One of the key components of the system is the path planning module, which combines the modified A* algorithm, insertion heuristics, and auction algorithm to calculate obstacle-avoiding routes for multiple responders with multiple destinations. A spatial data model is designed to support the storage of information about the tasks and routes produced during the disaster response. Our system has been validated using four navigation cases. Some preliminary results are presented in this paper and show the potential of the system for solving more navigation cases.

Introduction

First responders play an important role in crisis management, saving people's lives during disasters. They are personnel from different agencies, e.g., fire brigade, police and medical care, and are responsible for a wide range of tasks, including searching for survivors, transporting relief goods, and evacuating and transferring wounded people. Most of these emergency tasks require fast and safe navigation as well as the coordination of the response teams. As the economic and human losses due to natural, man-made and human invoked disasters are increasing (Munich, 2015), much more research efforts have been devoted to the issues in disaster management and a special attention has been paid to the navigation for first responders in disaster response.

One of the challenging issues is that natural or man-made disasters can create all sorts of moving obstacles that affect the road network, making some roads dangerous or impossible to traverse. Using traditional routing algorithms, Mioc et al. (2008) and Chitumalla et al. (2008) develop applications that use the forecast information of hazards in the near future in routing and providing navigation services taking blocked areas or streets into account. Nevertheless, they lack a consideration of the dynamics of the environment affected by the moving obstacles, which may make the planned path longer than the shortest one. In some situations, the responders can pass through the threatened roads before they are affected, instead of just avoiding them. Because the status of the road network affected by moving obstacles changes over time, the temporal aspect should also be considered in the routing process. Similar research in the field of robotics on navigation in the presence of moving obstacles has been considerable (Phillips and Likhachev, 2011, Li et al., 2009, Narayanan et al., 2012), which could be beneficial to the research on navigation for first responders in some aspects. Another important issue is that the collaborative activities among emergency agencies require the coordination of their routes and destinations. Because the first responders often work in groups and perform tasks together, they need not only to obtain individual routes but also to take into account other response units in the routing process. For example, in the case of emergency medical service, ambulances are distributed to different destinations to pick up and deliver patients according to factors such as the situation of the patients, the deployment of the paramedics, and the availability of medical supplies in hospitals. Although numerous techniques have been proposed in logistical planning and robotics to achieve the efficient allocation of vehicles (Dias et al., 2006), they do not have any effective mechanisms to deal with moving obstacles, and cannot be applied to the environment affected by disasters. Therefore, there is a great need for an emergency navigation system that is capable of quickly and safely navigating multiple responders to multiple destinations, avoiding moving obstacles.

In this study, we apply the agent technology to address navigation problems with moving obstacles. The method of agent technology was introduced by Wooldridge & Jennings (1995) and represents a diagram for the development of software entities that automates specific computer-based tasks. Agent technology has a set of features, including supporting distributed control, allowing for flexibility and adaptability, etc., which make it suitable for development of the system for navigation in the presence of moving obstacles (Wang & Zlatanova, 2013b). The agent technology has been applied to very varied fields, for example, crisis management (Schoenharl & Madey, 2011) and supply chains (Vok et al., 2010, Zeddini et al., 2008). Recently, there has been increasing interest in applying agent technology to GIScience (Sengupta and Sieber, 2007, Crooks and Wise, 2013). Most work in this direction has been devoted to the use of multi-agent systems to simulate and study the behavior and interactions of humans in environments (Chen and Zhan, 2008, Torrens et al., 2012, Bonabeau, 2002). On the other hand, more and more researchers have applied the agent technology to the development of software spatial agents, which assist users in the analysis and evaluation of spatial problems (Genc et al., 2013, Nourjou et al., 2013). However, these agents are usually developed to address a particular set of geospatial issues, which limits their application to other problems, such as emergency navigation. Because the path planning in the presence of moving obstacles is characterized by its reliance on the large amounts of dynamic spatial data produced in disasters (Visser, 2009, Mioc et al., 2008, Wang and Zlatanova, 2013a), special types of agents, combined with GIS functionalities, are needed to quickly process and analyze the spatio-temporal data needed for navigation during the disaster response.

In this paper, we focus on addressing a subset of navigation cases presented in Wang & Zlatanova (2013c), and propose an integrated navigation system for first responders in the presence of moving obstacles, based on a multi-agent system. Our system extends the framework described in Wang & Zlatanova (2013b), and introduces a set of modules that are comprised of software agents for its spatial data processing and analysis. Such an approach can allow the system to be easily adjusted and applied to various navigation cases. We use hazard models to provide the predicted information about the obstacles, and select a geo-database in which to store the data needed for emergency navigation. A new one-to-one path planning algorithm is adapted from Wang & Zlatanova (2013a), and is combined with other path planning algorithms to calculate obstacle-avoiding routes for responders with one or multiple destinations. The remainder of the paper is organized as follows: Section 2 presents the navigation cases we aim to address in this paper, and describes the conceptual framework designed for navigation in the presence of moving obstacles. In Section 3, we illustrate the architecture of the proposed multi-agent based navigation system, which contains different types of agents and geo-database. In Section 4, we provide the algorithms used for the different types of path planning problems. Section 5 describes the implementation of our proposed system, and presents some of the results of applying this system to four navigation cases. Finally, we discuss some aspects of the system and conclude with an outline of future research in Section 6.

Section snippets

Conceptual analysis and design

This section first provides a taxonomy of navigation in the presence of obstacles. In this taxonomy, we examine and analyze different navigation cases, and place our studies in this paper in the broader context of navigation for first responders. After that, we present our approach, which supports routing in these navigation cases, and describe the general architecture of our navigation system.

The architecture of the multi-agent based navigation system

In this paper, we concentrate on issues related to the multi-agent system and geo-database. In the following sections, we elaborate on the design and development of the agents (the definition of the roles, functionalities, interactions, etc.), routing algorithms, and a spatial data model that structures the data required for the routing and allocation process. Aspects regarding the hazard simulation models are outside the scope of this paper.

The architecture of the multi-agent based navigation

Algorithms used for route determination in the path planning module

Basically, the algorithms calculate optimal or near-optimal routes for three types of path panning problems: one-to-one, one-to-many, and many-to-many, considering multiple parameters (the time required to carry out the task, the speeds of the vehicles, and the departure times). The details of the algorithms follow.

Implementation and case studies

Following the structure of the system presented in the previous sections, a prototype of a multi-agent based navigation system has been implemented. In our system, we use JADE (Java Agent Development Framework) (Bellifemine et al., 2005) as the underlying agent infrastructure, and combine it with another agent-based toolkit, GeoMASON (Sullivan et al., 2010). GeoMASON is a GIS extension of the simulation toolkit Mason (Luke et al., 2004). The proposed spatial data model is realized in the

Conclusions and future research

In this paper, we presented a multi-agent based navigation system, with the aim of providing solutions to different types of navigation problems. The proposed system integrates hazard models, agent technology, GIS technology, a geo-database, and routing algorithms. We used hazard simulations to provide spatial–temporal information about moving obstacles. A set of software agents was designed and developed to support the spatial data processing and analysis involved in the routing process. To

References (30)

  • A.T. Crooks et al.

    GIS and agent-based models for humanitarian assistance

    Computers, Environment and Urban Systems

    (2013)
  • P.M. Torrens et al.

    An extensible simulation environment and movement metrics for testing walking behavior in agent-based models

    Computers, Environment and Urban Systems

    (2012)
  • Z. Wang et al.

    A data model for route planning in the case of forest fires

    Computers & Geosciences

    (2014)
  • F. Bellifemine et al.

    Jade — A java agent development framework

  • E. Bonabeau

    Agent-based modeling: Methods and techniques for simulating human systems

    Proceedings of the National Academy of Sciences

    (2002)
  • X. Chen et al.

    Agent-based modelling and simulation of urban evacuation: Relative effectiveness of simultaneous and staged evacuation strategies

    Journal of the Operational Research Society

    (2008)
  • P.K. Chitumalla et al.

    Emergency response applications: Dynamic plume modeling and real-time routing

    IEEE Internet Computing

    (2008)
  • M.B. Dias et al.

    Market-based multirobot coordination: A survey and analysis

    Proceedings of the IEEE

    (2006)
  • Z. Genc et al.

    Agent-based information infrastructure for disaster management

  • S. Koenig et al.

    The power of sequential single-item auctions for agent coordination

  • M.G. Lagoudakis et al.

    Auction-based multi-robot routing

  • H. Li et al.

    Neural-network-based path planning for a multirobot system with moving obstacles

    IEEE Transactions on Systems, Man, and Cybernetics Part C: Applications and Reviews

    (2009)
  • S. Luke et al.

    Mason: A new multi-agent simulation toolkit

  • D. Mioc et al.

    On-line street network analysis for flood evacuation planning

  • R.E. Munich

    Topics Geo: Natural catastrophes 2014: Analyses, assessments, positions

    (2015)
  • Cited by (31)

    • Distant object localization with a single image obtained from a smartphone in an urban environment

      2022, International Journal of Applied Earth Observation and Geoinformation
    • A multiagent systems with Petri Net approach for simulation of urban traffic networks

      2021, Computers, Environment and Urban Systems
      Citation Excerpt :

      The inter-agent communication depends on the physical features of the systems, for instance, the proximity of agents, the wireless communication coverage, etc. (Wang, Cao, Wang, & Alsaadi, 2019; Yang, Zhang, He, Han, & Peng, 2020). The MAS formalism has been useful to represent the coordination motion of mobile robots, large economic systems, project task planning, determination of optimal routes (Wang & Zlatanova, 2016), social networks, etc. An agent-based approach applied to traffic systems is proposed in (Liu, Baldi, Yu, & Chen, 2020) by partitioning a complex system into agents and the design of heuristic algorithm for solving the switch-based adaptive dynamic programming (ADP) in a distributed way.

    • Safe multi-cluster UAV continuum deformation coordination

      2019, Aerospace Science and Technology
      Citation Excerpt :

      Model predictive control (MPC) [42] is a well-known approach for trajectory (control vector) optimization that accounts for motion costs and vehicle dynamics constraints [43]. Researchers have proposed centralized [44–46] and decentralized [47–49] approaches for multi-agent path planning. Ref. [44] applies A* for first responder multi-agent team planning in a cluttered environment while Ref. [45] applies mixed integer programming.

    • DOME: Drone-assisted Monitoring of Emergent Events For Wildland Fire Resilience

      2023, ICCPS 2023 - Proceedings of the 2023 ACM/IEEE 14th International Conference on Cyber-Physical Systems with CPS-IoT Week 2023
    View all citing articles on Scopus
    View full text