Formal Language Methods Based Modeling for Traffic Incident Management

Traffic Incident Management is a multi-jurisdictional process. Complications with communication, compatibility, coordination, institutional responsibilities and legal issues are inherent in a traffic incident management system. Increasing delay in incident clearance due to various conflicts has vital economical, safety, environmental and social impacts. Therefore, a thorough and rigorous modeling of the system is necessary to better understand its properties and systematically solve issues that might arise. This paper proposes the use of formal language theory for modeling, analyzing, and implementing the traffic incident management process. This theory has been used very effectively for hardware and software systems. Using formal languages, allows us to perform debugging on a traffic incident management system covering all possibilities for inefficiencies and problems for which we can find solutions. This paper demonstrates how to use formal language methodology to model the traffic incident management system through a case study in the Las Vegas area. Shlayan, Kachroo, Gibby, Ohene 1 Glossary for Actions Used in FSP Models: Action Translation alt_route Alternative route anthrtow Another tow arriveloc Arrive to location call_rc call received callf Call fire department congt_clrd Congestion cleared congtn_clrd Congestion is cleared congtnotclr Congestion not cleared driveloc Drive to location ernotarrv Emergency responders have not arrived ernotreq Emergency responders are not required ertaskincmplt Emergency responders task is not complete eqptavl Equipment available fbusy Fire department is busy fmbusy Fire and Medical are busy freq Fire department is required mbusy Medical department is busy mreq Medical required noeqpt No equipment rdnt_call Redundant call tow_nformed Towing company is informed townotavl Towing truck not available trfcjam Traffic jam trfjam Traffic jam INTRODUCTION Traffic incidents are non-recurring events in traffic flow operations that often cause delay due to congestion and safety hazards (8). These incidents account for approximately one third of all delay on the US highways caused by traffic congestion and are responsible for nearly 60 percent of delay triggered by weather, construction, and special events (13). The operating capacity of a typical highway is reduced by 63 and 77 percent during one lane and two lane obstructions, respectively, on a three lane freeway segment (3). Incidents, such as a disabled passenger car parked on the shoulder of the roadway reduce the available capacity by up to 17 percent (3). In addition, crash reduction or avoidance can be significant as illustrated by an evaluation conducted by Minnesota Department of Transportation. This evaluation in 2004 reported that 68 percent of the monetary benefits of a traffic incident management program was a reduction in crashes. The impact of traffic incidents stretches beyond safety degradation and traffic congestion. Human productivity loss and fuel waste are definite economical outcomes (17). In 2005, congestion costs were estimated to be $78.2 billion in 437 U.S. urban areas where 52 to 58 percent of the total motorist delay is due to traffic incidents (3). When an incident occurs, various public agencies, such as ones related to medical, law enforcement, fire, and other public emergency agencies are usually among the first to respond. In addition, private agencies such as towing companies and hazardous materials contractors are most likely to be involved (13). On one hand, the existence of specialized entities delivers high quality work in handling tasks at the incident scene. On the other hand, this also raises challenges since each of these agencies has uniquely different priorities Shlayan, Kachroo, Gibby, Ohene 2 and views (13). Moreover, every agency has a separate communication system through which dispatchers communicate information about the incident to their agents. Such independency in communications leads to further misunderstanding among the various agents present at the scene leading to even more delays in the incident clearance process. Carson, et. al. (4), conducted a comprehensive evaluation of an incident response team program for the Washington State Department of Transportation determining its effectiveness. The study claims a 20.6 minute reduction in average duration of incidents from 1994 to 1995 resulting in $20,600 to $61,800 savings per incident (4). Therefore, an organized traffic incident management process promoting integration and bonding of multiagency operations and communications at the incident scene is necessary. A well planned incident management system, through formal and informal processes, improves efficiency and coordination between the multi-jurisdictional responses reducing incident clearance times and vehicle delays (17). However, such coordination is faced with many obstacles that are inherent in the system, such as uncertainty, sudden events, resource shortage, faulty information, and disruption of infrastructure support (5). In addition to support systems, incident management is currently formulated and implemented conventionally based on manual methods relying on personal experiences of the personnel from within the incident management field which has its shortcomings (7). The current conventional approach does not provide timely information on traffic networks, and it does not allow for conflict detection, or alternative incident management scenario evaluation due to time constraints (7). Furthermore, personal experiences are likely limited; they also vary from person to person’s experiences which is the main factor in the decision making process. This often leads to further conflicts and difficulties and, thus adversely impacts the traffic operations. In this paper, using formal languages methodology for modeling, analyzing, and implementing a unique Incident Management process is proposed. Using formal language theory, allows us to perform rigorous debugging on existing and future incident management systems covering all possibilities for inefficiencies and problems for which we can find solutions. The modeling approach introduced herein provides the flexibility to reflect on any Incident Management process depending on various variables involved for a certain urban region. Through formal methods modeling, customized software tools can be developed for a specific region enhancing the Incident Management process significantly. In section 2, a literature review on previous work for Incident Management modeling will be discussed. In section 3, the IM process in the Las Vegas area is described. An overview of the used modeling method and approach are presented in section 4. A demonstration on how formal methods are used to model the IM process through a case study is presented in section 5 followed by further analysis in section 6. Then, conclusions are in section 7. LITERATURE REVIEW Effective traffic incident management systems consist of three main aspects, multiagency communications and control, decision making, and sharing of limited resources. For a model to be successful these three aspects have to be addressed, otherwise complications in the incident management system may be overlooked. In this section, some proposed approaches to improve Incident Management processes are discussed. Incident prediction models based on analysis of incident patterns, frequency, and duration were proposed by Konduri, et. al. (10) and used for improving the freeway management system by assessing various IM strategies. Such models would be very useful in incident management systems; however, methods based on static data are not sufficient to comply with the required short term actions necessary in the most effective incident management systems (18). An agent based approach for monitoring, analyzing, and supporting Incident Management processes by error detection and providing support for such errors is proposed in (6). Temporal Trace Language (TTL) was used as a tool for formal representations of system’s properties. The author’s approach is adequate; however, the scope of this modeling is error detection for improving techniques in current incident management support systems that detect contradictory and unreliable information. This approach does not address broader issues in incident management such as the overall interaction and harmony between the involved Shlayan, Kachroo, Gibby, Ohene 3 agencies, limited and shared resources, or liveness properties of the incident management system as a whole. Mingwei in (7) proposes a real-time evaluation and decision support system (REDSS) for IM which detects traffic incidents, estimates impacts of incidents, formulates guidance scenarios, and monitors and evaluates scenario implementations. REDSS integrates a series of information analysis and processing technologies such as data fusion, expert systems, data warehousing and data mining (7). However, REDSS has not been validated. Chen in (5) , the author recognizes the constraints on responder’s capabilities to analyze coordination problems due to the requirement of rapidness in decision making. Therefore, a life cycle approach is introduced providing a broad and systematic view of activities relating to emergency response management. An inter-vehicular communication system design is proposed in (16). This system provides the ability to quickly discover and transmit real time multimedia information from an incident location to the approaching first responders (16). Kim in (9) introduces a conceptual model explaining the efficiency of decision-making of the Critical Incident Management Systems (CIMS). Researchers have demonstrated numerous attempts in improving the incident management process (7); however, in the history of IM, such attempts have been mainly focused on supporting systems. Such systems are mainly used to assist participating agency in assigning tasks and making decisions. Clearly, these systems do not integrate the three aspects necessary for a successful incident management modeling as well as implementation of strategies. A successful IM necessitate a broad and integrated response to incidents (14). In this paper, an incident management representation is proposed which provides the ability to account for all three aspects of IM integrating them into one systematic model. Moreover, specifying properties for the system and verifying them before implementation. The proposed model provides the ability to validate and verify any existing incident management process including supporting systems; most importantly, it provides a method to verify the interaction between such systems as well as multiagency processes INCIDENT MANAGEMENT IN THE LAS VEGAS AREA Until recently, Las Vegas (map depicted in Figure 1) has been one of the fastest growing cities in the United States. Consequently, highway capacity investment has not been able to keep pace with the growth in traffic; therefore, major roadways are experiencing substantial congestion during off-peak periods as well as peak periods. Users cost per hour for a closure on I-15 was recently estimated at $240,000 and can go up to $750,000 during the afternoon peak period (8). The report produced by Iteris (1) identified the existing institutional relationships which include operational agreements between various agencies for the Las Vegas area. Furthermore, it showed responsibilities of various organizations during an incident management process. Emergency responders in Las Vegas include but are not limited to the following agencies: Department of Safety Nevada Highway Patrol (NHP), Las Vegas Metropolitan Police Department (LVMPD), Regional Transportation Commission of Southern Nevada (RTC), Freeway Arterial Transportation System (FAST), Clark County Office of Emergency Management and Homeland Security, Clark County Fire Department (CCFD), and Coroner’s Office . A local traffic incident management (TIM) Coalition has been formed where various emergency responder agencies meet and discuss regional issues involving traffic incidents in the hopes of resolving any boundaries resulting in an improved communications, enhanced coordination, and an efficient incident management process in the Las Vegas area (1). FAST center operates the freeway and arterial traffic signal systems. FAST also supports incident management through traffic control (1). According to the Incident Management Strategies Draft Report (8), incident management is the key motivation for the existence of FAST in Las Vegas. Specifically, FAST provides data and tools to identify incidents and assists with remote monitoring of the incidents. Las Vegas has witnessed drastic improvement in the incident management process as a result of FAST efforts in detecting and monitoring incident occurrences, and the TIM team attempts to resolve any Shlayan, Kachroo, Gibby, Ohene 4 FIGURE 1 Las Vegas Map, freeway and arterials miscommunication issues between local agencies. However, crash data, presented in tables 1(a) and 1(b), from 2003 through 2008 obtained from the Computer Aided Dispatch (CAD) Center of LVMPD and a year worth of data from NHP were analyzed; as depicted in Figures 2(a) and 2(b). It was found that the average management and clearance times of incidents need improvement. Thus, a systematic solution through qualitative modeling is necessary for finding all possible sources of inefficiencies. MODEL APPROACH The purpose of formal languages theory is to bring order to complex system anarchy (15). Formal languages are characterized by predefined rules such as formal notations in mathematics, logic, and computer science (2, 15). A finite automaton is a string processor that assists in defining certain formal languages by accepting or rejecting a sequence of symbols (15); Applications that require pattern recognition techniques have fundamental interest in finite automata (2). A deterministic finite automaton consists of a finite number of states or conditions in which a system can exist and only one of these states can be an “initial” state. Additionally, Shlayan, Kachroo, Gibby, Ohene 5 TABLE 1 Average Arrival, Management, and Clearance times for incidents that occurred on the I15 and arterials in the Las Vegas area (a) LVMPD data Year AVG Arrival Time AVG Management AVG Clearance Time 2003 0:18:04 1:05:16 1:23:29 2004 0:22:37 1:11:20 1:33:39 2005 0:25:08 1:10:29 1:35:37 2006 0:25:09 1:13:12 1:38:21 2007 0:21:51 1:13:28 1:34:38 2008 0:19:47 1:43:21 1:46:00 (b) NHP data Month AVG Arrival Time AVG Management AVG Clearance Time Jul-08 0:11:30 1:12:04 1:30:01 Aug-08 0:11:10 1:13:22 1:30:27 Sep-08 0:11:06 1:13:18 1:31:35 Oct-08 0:10:53 1:16:39 1:33:47 Nov-08 0:10:43 1:08:20 1:25:59 Dec-08 0:13:21 1:10:58 1:31:33 Jan-09 0:11:22 1:13:12 1:28:49 Feb-09 0:12:12 1:08:40 1:27:29 Mar-09 0:11:42 1:07:50 1:24:54 Apr-09 0:12:38 1:08:42 1:27:07 May-09 0:11:22 1:07:06 1:23:33 Jun-09 0:11:52 1:07:18 1:26:40 such an automaton must contain at least one or more “terminal” or “accepting” states. Transitioning may be performed through two different actions, either switching to another state or remaining in the current state (2). Execution of state transition depends on the current state and the action or inaction identified by a symbol Using finite automata, a simple example of an incident may be modeled in a pictorial form called a state diagram, as depicted in Figure 3, glossary for FSP actions is introduced before the introduction. State “S0” represents a pre-accident situation which might imply traffic is in a free-flow state. The symbol “accident” represents an occurrence of an incident which causes the system to switch to state “S1” implying an incident scene. Once the system is in state “S1” only two transitions are possible represented by the symbols “call_911” and “callfailed”; first of which causes the system to switch to state “S2” implying that the incident is in the management process. The second symbol causes the system to remain in the same state implying that no advancements can be made unless an emergency responder is informed. Once the system reaches the management state, it can switch states when congestion is cleared (“congtn_clrd”) going back to free traffic flow in pre accident conditions (state “S0”). It remains in the management state “S2” if congestion is not cleared (“stillcongested”). Modeling the evolution of an incident scene using finite automata is methodically appropriate in terms of a sequence of events. Furthermore, many transitioning possibilities can be considered, depending on various conditions, which adds flexibility in modeling any Incident Management system. Every Incident Management process, however, is an interaction between multiple processes occurring concurrently. Thus, concurrency is an aspect that must be addressed in the Incident Management model. Shlayan, Kachroo, Gibby, Ohene 6 (a) LVMPD data, incidents occurred on arterials (b) NHP data, Incidents occurred on the I15 FIGURE 2 Average Arrival, Management, and Clearance times for incidents that occurred in the Las Vegas area Shared actions in Labeled Transition Systems (LTS) provide the ability to model concurrent finite state machine processes. They are described textually as finite state processes and displayed and analyzed by the LTS analysis tool (11). The LTS analysis tool provides the possibility to structure complex systems as sets of simple activities represented as sequential processes using Finite State Processes (FSP) (11). Processes can overlap or run concurrently reflecting real world situations as in the Incident Management process. Finite state processes (FSP) have a predefined language for their description; actions can be described using the action operator “ → ”. For instance, (x → P ) describes a process that initially engages in the action “x” and then behaves as described by process P (11). In order to model choice, the choice operator Shlayan, Kachroo, Gibby, Ohene 7 S1 S0 S2 callfailed accident call_911