Year: 2024
Type: editorial
Abstract: Introduction to the Special Issue on Cooperative Decision MakingArtificial intelligence (AI) technologies are continuing to be integrated in various forms of mobility.With this rapid emergence of advanced machine intelligence, a host of new challenges arises when multiple autonomous transportation systems need to cooperate to make decisions towards their goal of producing well-balanced and coordinated activities.This situation arises frequently in mobility domains, ranging from (semi-)autonomous vehicle fleets to smart traffic management systems, where agents need to infer the intentions of other mobile agents and reflect such information in their own decision making.The decentralized nature of the problem makes this inference nontrivial, especially where resources (space and time) are to be shared by multiple agents.Anomalous behavior of a certain agent could influence the behavior of other agents in an unexpected manner.The problem is even more complex when machines and humans must reach a consensus among their distributed decision-making processes.To meet this challenge, it is important to understand and explore fundamental algorithmic and cognitive theories and processes in cooperative decision making.The goal of this special issue is to bring together academic researchers and industrial professionals to address fundamental challenges in various aspects in machine intelligence and human cognitive modeling for cooperative decision making, focusing on underlying theories and how they are implemented in practical transportation systems.Paper (1) "Argument-based Cooperative Decision-Making for Autonomous Vehicles" addresses multi-criteria decision problems for autonomous vehicles in dynamic environments, focusing on cooperative decision-making.It introduces a formalism using defeasible preference rules to represent agents' varying priorities and trade-offs.The approach generates arguments based on these rules and uses an argumentation framework to resolve conflicts and explain decisions.The study includes a prototype implementation with interactive visual tools and demonstrates the method's adaptability to changing alternatives and evidence.This paper represents a formalized approach for cooperation that can handle the heterogeneity of mobile agents in transportation systems, supporting the ability to model distinct types of road users, such as passenger cars versus emergency vehicles, and resolve their preference conflicts as a component to handle the complexities found in real-world cooperative decision making.Paper (2) "Cooperative Advisory Residual Policies for Congestion Mitigation" addresses the problem of mitigating traffic congestion.This work advances learning-based advisory systems by introducing cooperative advisory residual policies that adapt to different driver behaviors.These policies are trained in simulation with models that address real-time driver responses and incorporate behavior inference to react to individual driving nuances.A novel reward function incentivizes advice that is easier for human drivers to follow.The policies are evaluated through a driving
DOI: 10.1145/3703933
Authors: Kikuo Fujimura, Mykel Kochenderfer, Kyle Wray
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4404998157