Luis A. García

Agent-Based management of non urban road meteorological incidents

Current protocols for defining and developing traffic strategies for dealing with non urban road meteorological incidents are human-centric based. These protocols are not easy to implement and, often, they do not work as well as proposed. This is mainly due to the following reasons: 1) traffic flow information is usually distributed in several equipments belonging to different traffic management centres, so the traffic operators only have a reduced view of the overall incident information; and 2) there are several traffic administrations and offices with competences in traffic management, so operators from different administrations and offices have to manually coordinated competences, resources and information. A Multiagent System (MAS) able to help traffic operators to determine the best traffic strategies and to help them in coordinating tasks is proposed in this paper. The A-3 Spanish freeway, a real traffic domain, is being used as the non urban domain where the proposed MAS is being evaluated.

A parallel algorithm for the two-dimensional cutting stock problem

Cutting Stock Problems (CSP) arise in many production industries where large stock sheets must be cut into smaller pieces. We present a parallel algorithm – based on Viswanathan and Bagchi algorithm (VB) – solving the Two-Dimensional Cutting Stock Problem (2DCSP). The algorithm guarantees the processing of best nodes first and does not introduce any redundant combinations – others than the already present in the sequential version. The improvement is orthogonal to any other sequential improvements. Computational results of an OpenMP implementation confirm the optimality of the algorithm. We also produce a new syntactic based reformulation of the 2DCSP problem which leads to a concise representation of the solutions. A highly efficient data structure to store subproblems is introduced.

Complexity Driven Performance Analysis

This work presents a new approach to the relation between theoretical complexity and performance analysis of MPI programs. The performance analysis is driven by the information produced during the complexity analysis stage and supported at the top level by a complexity analysis oriented language and at the bottom level by a special purpose statistical package.

A Distributed and Automatic Negotiation Protocol for Resource Allocation in the Management of Traffic Accidents

Several resources must be devoted to deal with the management of a traffic accident. However, these resources are managed by several independent agencies. The current solutions to coordinate the resources of these agencies are being provided manually by using centralized systems, obviating the distributed nature of the problem. This paper presents a new distributed combinatorial auction PAUSE based protocol suitable to be used in these types of situations. The main innovation in this new protocol is that it allows to allocate efficiently resources grouped by types. In addition it can exploit the existing synergy between the resources provided by any single agency. A multiagent system, MAS, has been programmed to evaluate the performance of this protocol in the traffic accidents domain. The assignments of resources proposed by this MAS show that the protocol provides efficient solutions close to the optimal ones.

An autonomic system for intelligent truck parking

Traffic and road transport conditions are strongly influenced by the decisions of drivers. The more information they receive, the better decisions may be taken about their behavior. This is especially important in the transport of goods, since drivers are subject to traffic laws about maximum driving time and minimum time rest. In this paper a multi-agent system for negotiated management of parking spaces in road rest areas is presented. This system dynamically adapts itself to the preferences and needs of the drivers of goods about parking requests. The system is shown to be robust to incidents regarding the closure of road rest areas and an increased volume of freight traffic. The results also show that the number of illegal parkings is reduced resulting in greater road safety.

Two-Dimensional Cutting Stock Problem: Shared Memory Parallelizations

Cutting stock problems arise in many industries where large stock sheets of a given material must be cut into smaller pieces. Many algorithms have been proposed for solving each of the problem formulations. We present different implementations based on Viswanathan and Bagchis algorithm to solve the two-dimensional cutting stock problem (2DCSP). One approximation parallelizes the generation of new builds from different subproblems. Also a highly efficient data structures to store subproblems are introduced, allowing to provide a parallel implementation where the generation of new subproblems from a particular one can be distributed. The OpenMP tool has been used for the parallel implementations and some computational results are presented

A Framework to Evaluate Autonomic Behaviours for Intelligent Truck Parking

This paper presents a multi-agent platform to evaluate different strategies to manage the negotiated management of parking spaces in road rest areas. The system dynamically adapts itself to the preferences and needs of the drivers of goods about parking requests. The system is shown to be robust to incidents regarding the closure of road rest areas and allows a conversational interaction of new parking requests through an Android based mobile application.

Applying the PAUSETA Protocol in Traffic Management Plans

The application of a traffic management plan (TMP) currently involves a manual negotiation between some agencies. Human operators from these agencies must negotiate with each other about who has to give each of the resources required by the TMP. This chapter proposes to use a new protocol, the Progressive Adaptive User Selection Environment by Type Agreements (PAUSETA) protocol, to support the autonomous deploying of a TMP. This protocol relies on a distributed combinatorial auction that preserves confidentiality and legal competences in the use of each resource in each agency. PAUSETA addresses the resources required by a TMP by type, that is, in the natural way that is given in a TMP, and it exhibits some self-* properties of an autonomic system. This protocol has been applied to a real traffic scenario to deal with a traffic incident. The analysis of the simulations performed show that PAUSETA provides results closely related to the ones obtained by a centralized system, thereby avoiding issues with regard to confidentiality about data or legal competences.

A Cooperative Information System for Managing Traffic Incidents with the PAUSETA Protocol

When a serious traffic incident happens several independent agencies must be coordinated in order to allocate the necessary resources to attend it. Currently, the coordination among these agencies is done manually, increasing the response time. In this paper is presented an information system suitable to be applied to reach agreements in a distributed and autonomous way among agencies about which an agency should give which resource to manage a traffic incident. This information system is guided by the PAUSETA protocol, which executes a distributed combinatorial auction for resource allocation. By using this protocol several issues around private information about resource features and availability and administrative and legal competences are individually managed by each agency. This information system has been implemented following a multi-agent software architecture. The prototype has been tested for a real traffic incident scenario in Castellon (Spain). The results obtained are consistent with respect to the ones it might provide an optimum centralized system. However this optimum centralized system is impossible to apply due to the inherent distributed nature of the problem addressed.

A rule-based multi-agent system for local traffic management

Road Traffic presents a high dynamism which makes necessary the development of traffic management and control strategies to improve traffic flows and more important, road safety. So it is needed the use of intelligent systems to support traffic organizations and road operators to cope with incidents. In this paper we introduce a local autonomous system for traffic management. This way, though there could be a breakdown in the communications between the local system and the TCC, the local system will be able to warn the road users in case of incidents. The system uses multiagent technology to work with the specific characteristics of traffic domain. The expert MAS system is ruled-based that implies each agent has a knowledge and a fact base in order to reach their objectives.