|Parallel Simulation Technology|
PIM/Paracell Application Notes
Software complexity killed the Baggage Handling System (BHS) at Denver International Airport. The complexity that characterizes applications such as this overwhelms conventional software engineering methodologies. Parallel Simulation Technology PIM and Paracell were designed to solve the problems of complex, real-time applications such as BHS.
PIM embedded BHS is capable of handling Denver airport baggage handling-class problems. Furthermore, PIM embedded BHS can achieve high throughput while preventing future problems from adversely effecting the baggage handling capabilites of the airport. A PIM embedded BHS gives customers a system with high availability and efficiency.
The PIM technology provides:
1. high performance software that works
2. a dramatically simpler software implementation
3. high intellegence software that is fail-safe
4. scalable systems that can start small, then grow big
Two types of baggage handling systems are reviewed: Discrete Processing, where baggage is processed individually, and Batch Processing, where baggage is processed in groups. The bottleneck on both systems are in information processing, not controllers, nor the mechanical part of the system. Discrete processing BHS are needed to sort items in timely manner. Batch processing BHS require a fast scheduler for smooth operation. Both bottlenecks benefit well from the PIM technology. The PIM technology adds value to BHS that no other competitors can match.
Software is becoming increasingly complex. The demand for more functionality far outstrips the progress in hardware, and more so, in software technology. Recently, a case that received much attention is the Baggage Handling System at the new Denver International Airport (DIA). The system was designed to handle 4,000 'telecarts' that carry bags at a speed of about 28 km/h, servicing 20 airlines along 32 km of track. Controls were handled by 64 personal computers hooked to 5,800 electric eyes, 315 radio receivers, 182 switches, and at least 60 bar code scanners. The result was a catastrophe. Telecarts piled up and jammed on tracks. Test parcels were ripped, dropped, and misrouted. The few that were properly routed took too long to reach their destination. Consequently, opening of the airport has been delayed for over a year. The cost of this failure has been huge: $500,000 a day, just to finance the bonds.
This BHS was developed by BAE Automated Systems Inc., of Dallas. BAE has a long history in the materials handling business, and has designed and built BHS for major carriers at airports around the world for over 10 years. As demand for addtional functionality increased, so did the lines of code required to implement BHS systems. Software became theprimary problem, and to rewrite the software would be overwhelming The alternative was to keep pushing till it broke. Unfortunately for the city of Denver, this happened at its new airport.
Conventional software approach are powerless against such a complex system. Pushed just a little beyond its specifications, and the system breaks. The complexity overwhelms the conventional system. Problems are hard to track, and much harder to solve. Often, solutions are passed over for patches that merely treats the symptom. The system is a growing house of cards.
Agent based methodology flies in the face of this conventional thinking. Complex problems are divided into small unique components interacting independently in a world shared by many. The interaction of individual units give rise to the system behavior, not specifically defined, but generally described. Autonomous agents provide an elegant approach to complex problems. Agent-based methodology pushes the command & control level down. At this level, problems are easily tracked and simple to solve. The legacy of treating symptoms with dubious patch work is replaced by a solid problem solver. Agent-based technology brings elegant approachable solutions to complex problems.
The agent based technology of PIM is ideal for registering, tracking, sorting, and scheduling. These are complex problems that require PIM technology. At the heart of the DIA problem, was the inability of BAE system to handle such a complex problem. This is where the PIM technology shines. Highly reliable mechatronics and cell/line controllers minimize downtime, and the agent based technology of the PIM group/superviosry controller optimizes uptime. Innovative technologies are combined to provide high availability baggage handling system. Maximum throughput is possible with the combination of reliable mechatronics and robust intelligence.
The function of the BHS is to deliver baggage from the point of input to the designated point of output. In general BHS consists of the input, the output, and the transport.
The input is where the baggage enters BHS, such as the check-in counters or curbside counters. Baggage checked in at these counters enter the BHS as a group from a single drop-off point. Baggage also enters the BHS from the unloading airplanes. There are four types of input in general.
1. check-in counters
2. bulk input from check-in counters outside the airport
3. bulk input from arriving flights
4. input from another BHS (baggages of connecting flights)
The output is where the baggage leaves the BHS, such as the baggage claim area. In general, there are three types of output.
1. baggage claim area carousels
2. drop off area for departing flights
3. output to another BHS (baggages of connecting flights)
Connecting the input and the output is the main transport system. It accepts incoming items from many input conveyors, and delivers to many outgoing conveyors or output carousels. The majority of the input is sent to a single output. However, the input is not ordered, and must be routed. To route items to their designated output, it is necessary to register and track items. Items are registered at the input, and tracked from this point. Baggage is tracked until each bag's destination is reached.
The BHS problem, can be described as 'moving items from point A to point B,' which can be divided into three parts: movement, tracking, and routing.
The 'Movement Problem' covers the process of baggage transportation, i.e., physical movment of an item from point A to point B. The 'Tracking Problem' covers the process of item identification. Registration puts the item in the system, and continuous monitoring of sensor information tracks the item. The complexity of the Tracking Problem changes with the BHS type: discrete or batch processing (see next section). The complexity of the Tracking Problem is directly proportional to the amount of information on the communication network, and therefore the system performance.
Solution of the 'Routing Problem' determines how a particular item reaches its destination. The complexity of this problem depends on the type, discrete or batch. In discrete processing, baggage must be sorted. In batch processing, items must be scheduled. Both routers are the bottleneck of the system. At Denver, the problem lies mainly in routing items. Because there are many factors that affect the performance of the router, it is also the most vulnerable part of the BHS.
The BHS is a one way system that transports items from the terminal to the airplane, or from the airplane to the terminal. Figure 1 illustrates two baggage handling systems, connected by a transport system that carries baggage for connecting flights.
Single or multiple airlines can be served by the above system layout. The difference lies in the number of check-in counters and airplane docking areas. Also, the physical length and area that the BHS covers increases with every additional airline. Domestic and international flights of a single airline may require separate BHS.
Cargo transport also increases the input and output points of the system, but it is outside the scope of this paper. A cargo transport company is equivalent to an airline that serves limited routes with large containers. Usually cargo transport companies, such as Federal Express, DHL, maintain separate terminals with separate cargo (baggage) handling systems.
Two BHD types are considered: discrete and batch processing. In discrete processing systems, baggage from check-in counters feed into the main transport system freely. Tracking and routing becomes a major problem, as there is no order to the travelling baggage. This system is analogous to automobile traffic on highway systems.
A batch processing system maintains gates that control baggage in groups. Baggage travels in batches. Efficient scheduling of batches becomes a major problem, as all must channel through the main transport system. This system is comparable to train traific on railroads.
The transport system is considered to be the single main artery, or the collecting conveyor, which is fed by many input conveyors from the check-in counters (see figure 2). In discrete processing system, baggage enters the system freely, at any time. At the end of the collecting conveyor is a sorter that directs baggage to respective output conveyors. Discrete processing type BHS resembles highway system where baggage corresponds to automobiles.
There are some advantages to the discrete processing system. Because baggage enters the system freely, and without wait, only a minimal buffer is needed at the input. Because baggage enters independently, baggage does not have to be scheduled, i.e., the system is flexible.
There are few disadvantages, also. Items enter independently, and have independent destinations. Each item must be sorted to the correct destination by the sorter. The sorter becomes the choke point. Items cannot travel faster than the processing speed of the sorter.
Only identified baggage can be sorted. System wide tracking is one way to identify items entering the sorter. Radio Frequency (RF) transponders and bar codes are other alternatives. For items to be identified, items must be registered. Registered items can then be tracked, compared, and identified. Registration, tracking, comparison, and validation, all contribute to the volume of network information that slows system performance.
Baggage can enter the system freely, but it enters into flowing traffic. An entering item must wait for clearance to avoid collision. A system of traffic control is required for the entering items. A small buffer is necessary to hold items waiting for clearance.
In batch processing sytem, baggage enters the system in batches. Gates located between the input, the output, and the transport system, control the flow of the baggage travelling in batches (see Figure 3). System scheduler determines the state of the gates. Opening the gate between an input and the collecting conveyor releases the batch into the conveyor. A batch processing type BHS resembles a railroad system where a batch corresponds to a train.
There are some advantages to grouping baggage in batches. Baggage is processed in groups rather then individually. This reduces the amount of information processed. For example, groups of baggage, instead of individual items, are tracked. Routing is simpler because groups are fewer than items. A simple system made of gates between the collecting and the output conveyor can manage the routing. Gates control eliminates traffic control for the incoming items. Routing is incoporated into scheduling.
The batch process type BHS has its disadvantages. Items cannot enter the system as freely as it can in the discrete system. A system schedule determines when items can enter the main conveyor. Because items must wait for its turn to enter the collecting conveyor, large input buffers are needed. These buffers must be large enough to accommodate the worst case scenario.
An efficient scheduler provides smooth operation, making the scheduler the bottleneck. The scheduler can make or break the system, much like the sorter of the discrete process system. The inflexibility of the batch processing type system, can be compensated for by a flexible scheduler.
Both types of BHS transfer systems consist of motion controls, motion generators, sensors, and monitors. In a discrete processing system, sortersand gates are added to the configuration. The tracking systems inlcude registration, sensors, tracking processors, and monitors. Routing is accomplished through the use of sorters, a scheduler and an interface to the tracking system.II.
Applying PIM and Paracell to this Solution
Agent-based systems are ideal for an application with the complexity
of a BHS. In a large airport handling hundreds of thousands of
pieces of baggage per day such as DIA or Chicago's O'Hare, these
systems become extremely complex. Traditional top-down approaches
to software development for such systems produce a software house
of cards that is inflexible, difficult to modify, and brittle.
Taking an agent-based approach as enabled by the PIM architecture and Paracell, code is developed for the individual "agents" within the system, such as the controllers, sensors, and actuators. In developing this code, the engineer is interested only in that agents' domain. All agents have a global view of the environment in which they operate, and communicate through this environment. Thus, there is no need to develop code that links one software agent to another.
Paracell code in a BHS application is easily developed and implemented by domain experts. Rather that articulating needs to a system developer and going through lengthy design reviews and acceptance testing, those most familiar with the application take an active part in its developemnt. This minimizes implementation time, simplifies training, and makes in-house support, troubleshooting, and enhancement of an application possible.
The PIM system is designed for real-time applications. Its' execution is systolic at a constant frame rate that can be varied by the engineer according to the needs of the application. In a BHS where timing is critical, it is unacceptable to have the system performance impacted by the amount of code, interrupts, or other events.
Parallel Simulation Technology LLC