Logistic concepts and processes

Next: Predictive Statistic Network Control (PSNC)

Logistic concepts and processes

The logistics of any transportation system must

minimize trip times.

maximize reliability.

optimize use of resources.

The logistics of MAIT must be able to handle the particular properties of a MAIT network:

The MAIT network is a high dimensional deterministic system with hazardous perturbations:
The aim of a global logistic scheme must be to insure that locally occurring hazards remain local.

The logistics of MAIT is multi-layered:
user $\Rightarrow$ cabin $\Rightarrow$ carrier $\Rightarrow$ track.
The aim of a global logistic scheme is to care of the demand and resource patterns of each layer.

The MAIT network has some particular properties that requires new logistic concepts:

On the one side, the MAIT network can grow to a size, where events are no more exactly predictable, in contrast with small-size driver-less transport systems. Furthermore, isolated, hazardous events (like the breakdown of a network branch) can spread like waves across the network and corrupt all previously made predictions. On the other hand, automated systems with a global control are much more predictable than for example road traffic, because the destination and path of each vehicle is usually known to the global management systems--from the beginning of the trip. Therefore, MAIT is a high dimensional deterministic system with hazardous perturbations. The aim of a logistic concept must be to render the network-operation immune to those perturbations, in the sense that local perturbations remain local. Otherwise the system risks to become chaotic.
the logistics of MAIT are multi-layered: there are users that order cabins, cabin managements that require carriers and carrier that require track capacity.

In the following, we propose a scheme that includes these aspects, in particular the problem of determinism and probability. In a first step, some technical aspects are considered as for example which module is doing what and which information are exchanged with other modules. In a second step we propose a logistic method called Predictive Statistic Network Control (PSNC) and describe how certain quantities are estimated and how certain decisions are made.

Main high level logistic processes

Allocation:
these processes are required

to estimate or reestimate the time range during which a specific resources will be used.

to allocate the specific resource, if available, using the allocation database of the track management (generation of scheduling data).

Execution:
management modules give orders to other management modules or resources to execute a defined action.

These processes are performed in alternation in order to update allocation and estimation (as real time approaches estimated time).

$\begin{center}\vbox{\input{fig_proc_highlevel.pstex_t} }\end{center}$

The idea behind the strict separation between execution of actions (cabin or carrier movements, transfers, e.t.c) and allocation of resources (track capacity, empty carriers, cabins, e.t.c) is twofold:

each action that happens at present will necessarily influence the occupation of resources in the future. In case the present action cannot be performed exactly as planned, we need to reestimate the future of the system and, if required, to reallocate resources. Considering a fixed time instance in the future, we can improve our knowledge about the occupation of resources as we approach it in time, since the nearer future is more predictable. With other words, it is of advantage to update the estimated required resources in the future after each action in the present.
the allocation and execution are not always done in the same management module, because it often happens that only the amount of future resources is known but not exactly which module will be used, and which management is in control of this module. For example, if a cabin on carrier type A will change to carrier type B at a carrier exchange point. Then it is only necessary to assure that there are carriers of type B available as soon as the cabin arrives. Only at the moment when the cabin transfer takes place, it is certain which carrier is used to continue and which carrier management is involved to organize the future trip. Theoretically an exact resource allocation is possible, but it would be too restrictive for its technical realization and therefore sometimes impractical too expensive or too bulky. For examples simple queues as carrier stores would not be possible, but only random access facilities. Furthermore, an exact resource allocation would also be less tolerant to system failures or other unpredictable events.

Subsequently, we propose a scheme that takes care of the above mentioned issues and explain how, where and when the prediction and execution processes happen in a MAIT network.

Processes of allocation and prediction

module process global communication with

user management search/allocation of empty cabin track management

cabin management search/allocation of empty carrier track management

point-to-point routing^*

allocation of track capacity track management

carrier management if without cabin: point-to-point routing^*

if without cabin: allocation of track capacity track management

track management refined routing^** all other managements

time estimation

updating of allocation db

^* routing from and to stops and track cluster boarder-points.

^** exact routing of path inside cluster.

Example of track allocation/prediction cycle

$\begin{center}\vbox{\input{fig_cluster_alloc1.pstex_t} }\end{center}$

Example of track allocation/prediction cycle

$\begin{center}\vbox{\input{fig_proc_alloc.pstex_t} }\end{center}$

In this example the resources of the cabin are estimated and allocated. This means it is known at which point of the track the cabin will be and at what time (and at what probability as it is shown later). Thus, the track managements have to take care that the required resources are actually available at that time. For example track management 2 needs to open resources at the carrier queue of the exchanger in order to ``attract'' carriers for the estimated time. The entire chain of involved processes is quite large and is not shown in the example. But the allocation principles of carriers work in the same way as those of cabins. Note that with the cabin allocation the path on the track is known, but not exactly which carrier will be used after the cabin transfer at the carrier exchanger. This will be known during the execution at t₂ when the transfer actually happens.

Processes of execution

module process global communication with

user management informs user of trip user terminal

gives order to transport user cabin management(s)

cabin management gives order to load cabin cabin

put cabin in move mode cabin

supervises transfer cabin

gives order to transport user carrier management(s)

carrier management gives order to transport cabin carrier

supervises movement and transfer carrier

gives order to transport carrier track management(s)

track management gives routing tables for carriers track elements

Example of track execution cycles

$\begin{center}\vbox{\input{fig_cluster_exec.pstex_t} }\end{center}$

Example of track execution cycles

$\begin{center}\vbox{\input{fig_proc_exec.pstex_t} }\end{center}$

Note that carrier management₂ is not contacted until time t₂, when the transfer takes place and it is know which carrier is used to move cabin₁ ahead.

Next: Predictive Statistic Network Control (PSNC)

MAIT home Last updated: 2001-05-30 by
webmaster@maitint.org