Introduction to the Special Section on Selected Papers from ICCPS 2021

The article by Yuan et al. introduces an automated decentralized negotiation framework to resolve conflicts in smart city applications. This is an important research problem as conflicts across services directly affect users’ mobility and health in modern cities. The proposed framework of this article, called DeResolver, can achieve Pareto-Optimal agreement in an automated setting with limited additional overheads. DeResolver uses a two-step self-supervised learning-based algorithm to predict acceptable proposals and their rankings of each opponent through the negotiation. The proposed framework is evaluated on three different use cases (intelligent traffic light control, pedestrian service, and environmental control) in a non-trivial scenario and using real data. In this case study, the authors performed a data-driven evaluation of DeResolver using a large dataset consisting of the GPS locations of 246 surveillance cameras and an automatic traffic monitoring system with more than 3 million records per day to extract real-world vehicle routes. Through the case study analysis, the authors show that their framework can achieve a more equitable solution compared to a priority-based solution. Various performance metrics have been discussed and provided in the article.

Monitoring is of growing popularity as a lightweight tool for formal assurances, because it can often provide guarantees over short-term time horizons, yet is computationally tractable. The article by M. Waga et al. focuses on challenges that arise in monitoring due to sampling. This is particularly critical in networked CPSs, in which sampling intervals may be unduly long due to energy conservation or other constraints. The authors introduce a model-bounded monitoring scheme, in which prior knowledge about the system, framed in terms of linear hybrid automata (LHAs), is used to prune interpolation candidates between samples. Key to this is a novel description of a monitored language for LHAs that accommodates mode switching between samples. The authors show that the language membership problem can be reduced to an LHA reachability problem, which is readily solvable with state-of-the-art tools.

Sheng et al. present a trust-based route planner for automated vehicles that is validated in a driving simulator via a human subject experiment. The main contribution of the article is that in contrast to typical planning, the authors’ approach incorporates a data-driven model of the dynamics of the operator's trust in the automated vehicle and its planning capabilities. Such an approach is important for improving interaction between the automated vehicle and the human operator, because trust can influence not only whether the human operator wishes to take over but also their overall satisfaction with the automated vehicle. The authors construct a trust-aware planner that solves a POMDP that incorporates trust as a partially observable variable. In comparison to a planner that presumes a constant, pre-determined level of trust, the authors’ trust-aware planner results in fewer take-overs and higher operator satisfaction. The authors’ approach exploits structure in the POMDP dynamics that captures the cascading interactions of the vehicle's position, the type of hazard the vehicle encounters, and the anticipated success of the automated vehicle to manage the hazard.

The article by Pettet et al. presents a hierarchical dynamic resource allocation problem in a large, city-scale CPS. The authors model the allocation under uncertainty using Markov decision processes and address the scalability issues of MDP-based problem formulations and solutions. As a case study analysis, the authors use emergency response systems where they generate a sequence of decisions (e.g., assign ambulance services to an incident site) under uncertainty. The proposed methods are evaluated with real-world emergency response data from Nashville, TN.