Multi-agent Approach to Resource Allocation in Autonomous Vehicle Fleets

V2v Communication ◽

Solution Methods ◽

Multi Agent

The development of autonomous vehicles, capable of peer-to-peer communication, as well as the interest in on-demand solutions, are the primary motivations for this study. In the absence of central control, we are interested in forming a fleet of autonomous vehicles capable of responding to city-scale travel demands. Typically, this problem is solved centrally; this implies that the vehicles have continuous access to a dispatching portal. However, such access to such a global switching infrastructure (for data collection and order delivery) is costly and represents a critical bottleneck. The idea is to use low-cost vehicle-to-vehicle (V2V) communication technologies to coordinate vehicles without a global communication infrastructure. We propose to model the different aspects of decision and optimization problems related to this more general problem. After modeling these problems, the question arises as to the choice of centralized and decentralized solution methods. Methodologically, we explore the directions and compare the performance of distributed constraint optimization techniques (DCOP), self-organized multi-agent techniques, market-based approaches, and centralized operations research solutions.

Multi-Context System for Optimization Problems

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v33i01.33012929 ◽

2019 ◽

Vol 33 ◽

pp. 2929-2937

Author(s):

Tiep Le ◽

Tran Cao Son ◽

Enrico Pontelli

Keyword(s):

Complexity Analysis ◽

Optimization Problems ◽

Distributed Optimization ◽

Multi Agent System ◽

Constraint Optimization ◽

Agent System ◽

Multi Agent ◽

Definition Of ◽

This paper proposes Multi-context System for Optimization Problems (MCS-OP) by introducing conditional costassignment bridge rules to Multi-context Systems (MCS). This novel feature facilitates the definition of a preorder among equilibria, based on the total incurred cost of applied bridge rules. As an application of MCS-OP, the paper describes how MCS-OP can be used in modeling Distributed Constraint Optimization Problems (DCOP), a prominent class of distributed optimization problems that is frequently employed in multi-agent system (MAS) research. The paper shows, by means of an example, that MCS-OP is more expressive than DCOP, and hence, could potentially be useful in modeling distributed optimization problems which cannot be easily dealt with using DCOPs. It also contains a complexity analysis of MCS-OP.

A Two-Phase Complete Algorithm for Multi-Objective Distributed Constraint Optimization

Journal of Advanced Computational Intelligence and Intelligent Informatics ◽

10.20965/jaciii.2014.p0573 ◽

2014 ◽

Vol 18 (4) ◽

pp. 573-580

Author(s):

Alexandre Medi ◽

◽

Tenda Okimoto ◽

Katsumi Inoue ◽

◽

...

Keyword(s):

Branch And Bound ◽

Pareto Front ◽

Optimization Problem ◽

Optimization Problems ◽

Branch And Bound Algorithm ◽

Constraint Optimization ◽

Distributed Constraint Optimization Problem

Multi Objective ◽

Multi Agent ◽

A Distributed Constraint Optimization Problem (DCOP) is a fundamental problem that can formalize various applications related to multi-agent cooperation. Many application problems in multi-agent systems can be formalized as DCOPs. However, many real world optimization problems involve multiple criteria that should be considered separately and optimized simultaneously. A Multi-Objective Distributed Constraint Optimization Problem (MO-DCOP) is an extension of a mono-objective DCOP. Compared to DCOPs, there exists few works on MO-DCOPs. In this paper, we develop a novel complete algorithm for solving an MO-DCOP. This algorithm utilizes a widely used method called Pareto Local Search (PLS) to generate an approximation of the Pareto front. Then, the obtained information is used to guide the search thresholds in a Branch and Bound algorithm. In the evaluations, we evaluate the runtime of our algorithm and show empirically that using a Pareto front approximation obtained by a PLS algorithm allows to significantly speed-up the search in a Branch and Bound algorithm.

GPS-INS Integration Algorithm for Autonomous Vehicle Using Loosely Coupled Integration Architecture

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.390.506 ◽

2013 ◽

Vol 390 ◽

pp. 506-511

Author(s):

Rashid Iqbal ◽

Zhong Jian Li ◽

Khan Badshah

Keyword(s):

Autonomous Vehicles ◽

Low Cost ◽

Autonomous Vehicle ◽

Integrated System ◽

Measurement Unit ◽

Accurate Estimation ◽

Initial Alignment ◽

Integration Algorithm ◽

Loosely Coupled ◽

Integration Architecture

Inertial measurement unit (IMU) has been widely used for autonomous vehicles navigation. The accuracy of IMU specifies the performance of the inertial navigation system (INS).The errors in the INS are mainly due to the IMU inaccuracies, initial alignment, computational errors and approximations in the system equations. These errors are further integrated over time due to the dead-reckoning nature of the INS, which leads to unacceptable results. These errors need an accurate estimation for high precision navigation. INS is integrated with Global Positioning System (GPS) to estimate the errors and enhance the navigation capability of the INS. Linearized Kalman Filter (LKF) is proposed for estimating the errors in the low cost INS using Loosely Coupled integration approach, which is opted for its simplicity and robustness. Prediction part of the LKF is used during the GPS lag for errors estimation, which is found very effective for low cost sensors. The resulting GPS-INS integration algorithm is evaluated on simulated Autonomous vehicle trajectory, generated from 6-DOF model. The integrated system limits the attitude errors less than 0.1 deg and velocity errors of the order of 0.003 meter per second. Furthermore, it provides an optimal navigation solution than can be achieved from individual systems.

Distributed Constraint Optimization Problems and Applications: A Survey

Journal of Artificial Intelligence Research ◽

10.1613/jair.5565 ◽

2018 ◽

Vol 61 ◽

pp. 623-698 ◽

Cited By ~ 29

Author(s):

Ferdinando Fioretto ◽

Enrico Pontelli ◽

William Yeoh

Keyword(s):

Autonomous Agents ◽

Optimization Problems ◽

Constraint Optimization ◽

Uncertain Environments ◽

Agent Model ◽

Autonomous Behavior ◽

Multi Agent ◽

The field of multi-agent system (MAS) is an active area of research within artificial intelligence, with an increasingly important impact in industrial and other real-world applications. In a MAS, autonomous agents interact to pursue personal interests and/or to achieve common objectives. Distributed Constraint Optimization Problems (DCOPs) have emerged as a prominent agent model to govern the agents' autonomous behavior, where both algorithms and communication models are driven by the structure of the specific problem. During the last decade, several extensions to the DCOP model have been proposed to enable support of MAS in complex, real-time, and uncertain environments. This survey provides an overview of the DCOP model, offering a classification of its multiple extensions and addressing both resolution methods and applications that find a natural mapping within each class of DCOPs. The proposed classification suggests several future perspectives for DCOP extensions and identifies challenges in the design of efficient resolution algorithms, possibly through the adaptation of strategies from different areas.

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence ◽

Neural Regret-Matching for Distributed Constraint Optimization Problems

10.24963/ijcai.2021/21 ◽

2021 ◽

Author(s):

Yanchen Deng ◽

Runsheng Yu ◽

Xinrun Wang ◽

Bo An

Keyword(s):

Large Scale ◽

Optimization Problems ◽

Scale Up ◽

Constraint Optimization ◽

Confidence Bounds ◽

Multi Agent ◽

Constraint Optimization Problems ◽

Regret Bound ◽

Agent Coordination

Distributed constraint optimization problems (DCOPs) are a powerful model for multi-agent coordination and optimization, where information and controls are distributed among multiple agents by nature. Sampling-based algorithms are important incomplete techniques for solving medium-scale DCOPs. However, they use tables to exactly store all the information (e.g., costs, confidence bounds) to facilitate sampling, which limits their scalability. This paper tackles the limitation by incorporating deep neural networks in solving DCOPs for the first time and presents a neural-based sampling scheme built upon regret-matching. In the algorithm, each agent trains a neural network to approximate the regret related to its local problem and performs sampling according to the estimated regret. Furthermore, to ensure exploration we propose a regret rounding scheme that rounds small regret values to positive numbers. We theoretically show the regret bound of our algorithm and extensive evaluations indicate that our algorithm can scale up to large-scale DCOPs and significantly outperform the state-of-the-art methods.

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies ◽

SwiVR-Car-Seat

10.1145/3494968 ◽

2021 ◽

Vol 5 (4) ◽

pp. 1-26

Author(s):

Mark Colley ◽

Pascal Jansen ◽

Enrico Rukzio ◽

Jan Gugenheimer

Keyword(s):

Interface Design ◽

Autonomous Vehicles ◽

User Study ◽

Low Cost ◽

Autonomous Vehicle ◽

User Input ◽

The Road ◽

Car Seat ◽

Perceived Realism ◽

Vehicle Motion

Autonomous vehicles provide new input modalities to improve interaction with in-vehicle information systems. However, due to the road and driving conditions, the user input can be perturbed, resulting in reduced interaction quality. One challenge is assessing the vehicle motion effects on the interaction without an expensive high-fidelity simulator or a real vehicle. This work presents SwiVR-Car-Seat, a low-cost swivel seat to simulate vehicle motion using rotation. In an exploratory user study (N=18), participants sat in a virtual autonomous vehicle and performed interaction tasks using the input modalities touch, gesture, gaze, or speech. Results show that the simulation increased the perceived realism of vehicle motion in virtual reality and the feeling of presence. Task performance was not influenced uniformly across modalities; gesture and gaze were negatively affected while there was little impact on touch and speech. The findings can advise automotive user interface design to mitigate the adverse effects of vehicle motion on the interaction.

Risk Analysis of Autonomous Vehicles in Mixed Traffic Streams

Transportation Research Record Journal of the Transportation Research Board ◽

10.3141/2625-06 ◽

2017 ◽

Vol 2625 (1) ◽

pp. 51-61 ◽

Cited By ~ 23

Author(s):

Parth Bhavsar ◽

Plaban Das ◽

Matthew Paugh ◽

Kakan Dey ◽

Mashrur Chowdhury

Keyword(s):

Traffic Safety ◽

Failure Probability ◽

Autonomous Vehicles ◽

Autonomous Vehicle ◽

Fault Tree ◽

Communication Technologies ◽

Transportation Infrastructure ◽

New Combination ◽

Motor Vehicles ◽

Mixed Traffic

The introduction of autonomous vehicles in the surface transportation system could improve traffic safety and reduce traffic congestion and negative environmental effects. Although the continuous evolution in computing, sensing, and communication technologies can improve the performance of autonomous vehicles, the new combination of autonomous automotive and electronic communication technologies will present new challenges, such as interaction with other nonautonomous vehicles, which must be addressed before implementation. The objective of this study was to identify the risks associated with the failure of an autonomous vehicle in mixed traffic streams. To identify the risks, the autonomous vehicle system was first disassembled into vehicular components and transportation infrastructure components, and then a fault tree model was developed for each system. The failure probabilities of each component were estimated by reviewing the published literature and publicly available data sources. This analysis resulted in a failure probability of about 14% resulting from a sequential failure of the autonomous vehicular components alone in the vehicle’s lifetime, particularly the components responsible for automation. After the failure probability of autonomous vehicle components was combined with the failure probability of transportation infrastructure components, an overall failure probability related to vehicular or infrastructure components was found: 158 per 1 million mi of travel. The most critical combination of events that could lead to failure of autonomous vehicles, known as minimal cut-sets, was also identified. Finally, the results of fault tree analysis were compared with real-world data available from the California Department of Motor Vehicles autonomous vehicle testing records.

Solving Highly Cyclic Distributed Optimization Problems Without Busting the Bank: A Decimation-based Approach

Logic Journal of IGPL ◽

10.1093/jigpal/jzaa069 ◽

2020 ◽

Author(s):

Jesús Cerquides ◽

Juan Antonio Rodríguez-Aguilar ◽

Rémi Emonet ◽

Gauthier Picard

Keyword(s):

Belief Propagation ◽

Optimization Problems ◽

State Of The Art ◽

Empirical Evaluation ◽

The Novel ◽

Constraint Optimization ◽

Inference Algorithms ◽

Solution Methods ◽

Abstract In the context of solving large distributed constraint optimization problems, belief-propagation and incomplete inference algorithms are candidates of choice. However, in general, when the problem structure is very cyclic, these solution methods suffer from bad performance, due to non-convergence and many exchanged messages. As to improve performances of the MaxSum inference algorithm when solving cyclic constraint optimization problems, we propose here to take inspiration from the belief-propagation-guided decimation used to solve sparse random graphs ($k$-satisfiability). We propose the novel DeciMaxSum method, which is parameterized in terms of policies to decide when to trigger decimation, which variables to decimate and which values to assign to decimated variables. Based on an empirical evaluation on a classical constraint optimization benchmarks (graph coloring, random graph and Ising model), some of these combinations of policies, using periodic decimation, cycle detection-based decimation, parallel and non parallel decimation, random or deterministic variable selection and deterministic or random sampling for value selection, outperform state-of-the-art competitors in many settings.

A Connected Autonomous Vehicle Emulator (CAVE) for Testing Multi-agent, Conventional–Autonomous Mixed Vehicle Traffic Scenarios

Advances in Computers and Information in Engineering Research, Volume 2 ◽

10.1115/1.862025_ch11 ◽

2021 ◽

pp. 339-358

Author(s):

Dan Negrut ◽

Asher Elmquist ◽

Radu Serban ◽

Dylan Hatch ◽

Parmesh Ramanathan

Keyword(s):

Path Planning ◽

Autonomous Vehicles ◽

Autonomous Vehicle ◽

Computer Programs ◽

Software Infrastructure ◽

Vehicle Traffic ◽

Traffic Conditions ◽

Multi Agent

We discuss a software infrastructure that provides a virtual proving ground for designing, training, and auditing the computer programs used to pilot connected autonomous vehicles (CAVs). This effort does not concentrate on developing the piloting computer programs (PCPs) responsible for path planning in autonomous vehicles (AVs). Instead, we have established a first version of an emulation platform that changes the PCP design/test/improve process, which is often times carried out covertly [46], or in actual traffic conditions with potentially fatal consequences [45, 47].

Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence ◽

Learning Optimal Temperature Region for Solving Mixed Integer Functional DCOPs

10.24963/ijcai.2020/38 ◽

2020 ◽

Author(s):

Saaduddin Mahmud ◽

Md. Mosaddek Khan ◽

Moumita Choudhury ◽

Long Tran-Thanh ◽

Nicholas R. Jennings

Keyword(s):

Optimization Problems ◽

Optimal Parameter ◽

Exact Algorithms ◽

Mixed Integer ◽

Continuous Variables ◽

Multi Agent Systems ◽

Parameter Configuration ◽

Multi Agent ◽

Distributed Constraint Optimization Problems (DCOPs) are an important framework for modeling coordinated decision-making problems in multi-agent systems with a set of discrete variables. Later works have extended DCOPs to model problems with a set of continuous variables, named Functional DCOPs (F-DCOPs). In this paper, we combine both of these frameworks into the Mixed Integer Functional DCOP (MIF-DCOP) framework that can deal with problems regardless of their variables' type. We then propose a novel algorithm - Distributed Parallel Simulated Annealing (DPSA), where agents cooperatively learn the optimal parameter configuration for the algorithm while also solving the given problem using the learned knowledge. Finally, we empirically evaluate our approach in DCOP, F-DCOP, and MIF-DCOP settings and show that DPSA produces solutions of significantly better quality than the state-of-the-art non-exact algorithms in their corresponding settings.