| 1. |
- Yi, Yuhao, et al.
(författare)
-
Near-Optimal Resilient Aggregation Rules for Distributed Learning Using 1-Center and 1-Mean Clustering with Outliers
- 2024
-
Ingår i: Proceedings of the 38th AAAI Conference on Artificial Intelligence. - : Association for the Advancement of Artificial Intelligence (AAAI). ; , s. 16469-16477
-
Konferensbidrag (refereegranskat)abstract
- Byzantine machine learning has garnered considerable attention in light of the unpredictable faults that can occur in large-scale distributed learning systems. The key to secure resilience against Byzantine machines in distributed learning is resilient aggregation mechanisms. Although abundant resilient aggregation rules have been proposed, they are designed in ad-hoc manners, imposing extra barriers on comparing, analyzing, and improving the rules across performance criteria. This paper studies near-optimal aggregation rules using clustering in the presence of outliers. Our outlier-robust clustering approach utilizes geometric properties of the update vectors provided by workers. Our analysis show that constant approximations to the 1-center and 1-mean clustering problems with outliers provide near-optimal resilient aggregators for metric-based criteria, which have been proven to be crucial in the homogeneous and heterogeneous cases respectively. In addition, we discuss two contradicting types of attacks under which no single aggregation rule is guaranteed to improve upon the naive average. Based on the discussion, we propose a two-phase resilient aggregation framework. We run experiments for image classification using a non-convex loss function. The proposed algorithms outperform previously known aggregation rules by a large margin with both homogeneous and heterogeneous data distributions among non-faulty workers. Code and appendix are available at https://github.com/jerry907/AAAI24-RASHB.
|
|
| 2. |
- Besinovic, Nikola, et al.
(författare)
-
Artificial Intelligence in Railway Transport : Taxonomy, Regulations, and Applications
- 2022
-
Ingår i: IEEE transactions on intelligent transportation systems (Print). - : IEEE. - 1524-9050 .- 1558-0016. ; 23:9, s. 14011-14024
-
Tidskriftsartikel (refereegranskat)abstract
- Artificial Intelligence (AI) is becoming pervasive in most engineering domains, and railway transport is no exception. However, due to the plethora of different new terms and meanings associated with them, there is a risk that railway practitioners, as several other categories, will get lost in those ambiguities and fuzzy boundaries, and hence fail to catch the real opportunities and potential of machine learning, artificial vision, and big data analytics, just to name a few of the most promising approaches connected to AI. The scope of this paper is to introduce the basic concepts and possible applications of AI to railway academics and practitioners. To that aim, this paper presents a structured taxonomy to guide researchers and practitioners to understand AI techniques, research fields, disciplines, and applications, both in general terms and in close connection with railway applications such as autonomous driving, maintenance, and traffic management. The important aspects of ethics and explainability of AI in railways are also introduced. The connection between AI concepts and railway subdomains has been supported by relevant research addressing existing and planned applications in order to provide some pointers to promising directions.
|
|
| 3. |
- De Donato, Lorenzo, et al.
(författare)
-
Artificial intelligence in railways : current applications, challenges, and ongoing research
- 2023
-
Ingår i: Handbook on Artificial Intelligence and Transport. - : Edward Elgar Publishing. - 9781803929538 - 9781803929545 ; , s. 249-283
-
Bokkapitel (refereegranskat)abstract
- This chapter presents applications, challenges, and opportunities for the integration of artificial intelligence in rail transport, based on the current results of the European project Roadmaps for AI integration in the rail sector (RAILS). Past and ongoing research directions are briefly outlined, and then the regulatory landscape is presented as well as the main barriers to overcome. Some technical aspects are addressed to provide some valuable references, and a high-level description of ongoing research work is given, spanning from innovative studies on smart maintenance, collision avoidance, delay prediction, and incident attribution analysis to visionary scenarios such as intelligent control and virtual coupling.
|
|
| 4. |
- De Donato, Lorenzo, et al.
(författare)
-
Artificial intelligence in railways : Current applications, challenges, and ongoing research
- 2023
-
Ingår i: Handbook on Artificial Intelligence and Transport. - : Edward Elgar Publishing. - 9781803929545 - 9781803929538 ; , s. 249-283
-
Bokkapitel (refereegranskat)abstract
- This chapter presents applications, challenges, and opportunities for the integration of artificial intelligence in rail transport, based on the current results of the European project Roadmaps for AI integration in the rail sector (RAILS). Past and ongoing research directions are briefly outlined, and then the regulatory landscape is presented as well as the main barriers to overcome. Some technical aspects are addressed to provide some valuable references, and a high-level description of ongoing research work is given, spanning from innovative studies on smart maintenance, collision avoidance, delay prediction, and incident attribution analysis to visionary scenarios such as intelligent control and virtual coupling.
|
|
| 5. |
|
|
| 6. |
- Flötteröd, Gunnar, et al.
(författare)
-
Disaggregate path flow estimation in an iterated DTA microsimulation
- 2010
-
Rapport (övrigt vetenskapligt/konstnärligt)abstract
- This text describes the first application of a novel path flow and origin/destination(OD) matrix estimator for iterated dynamic traffic assignment (DTA) microsimulations.The presented approach, which operates on a trip-based demand representation,is derived from an agent-based DTA calibration methodology that relies onan activity-based demand model. The objective of this work is to demonstrate thetransferability of the agent-based approach to the more widely used OD matrixbaseddemand representation.The calibration (i) operates at the same disaggregate level as the microsimulationand (ii) has drastic computational advantages over usual OD matrix estimators inthat the demand adjustments are conducted within the iterative loop of the DTAmicrosimulation, which results in a running time of the calibration that is in thesame order of magnitude as a plain simulation. We describe an application ofthis methodology to the trip-based DRACULA microsimulation and present anillustrative example that clarifies its capabilities.
|
|
| 7. |
- Flötteröd, Gunnar, et al.
(författare)
-
Disaggregate Path Flow Estimation in an Iterated Dynamic Traffic Assignment Microsimulation
- 2014
-
Ingår i: Journal of Intelligent Transportation Systems / Taylor & Francis. - : Informa UK Limited. - 1547-2450 .- 1547-2442. ; 18:2, s. 204-214
-
Tidskriftsartikel (refereegranskat)abstract
- This article describes the first application of a novel path flow and origin/destination (OD) matrix estimator for iterated dynamic traffic assignment (DTA) microsimulations. The presented approach, which operates on a trip-based demand representation, is derived from an agent-based DTA calibration methodology that relies on an activity-based demand model (Flotterod, Bierlaire, & Nagel, 2011). The objective of this work is to demonstrate the transferability of the agent-based approach to the more widely used OD matrix-based demand representation. The calibration (i) operates at the same disaggregate level as the microsimulation and (ii) has drastic computational advantages over conventional OD matrix estimators in that the demand adjustments are conducted within the iterative loop of the DTA microsimulation, which results in a running time of the calibration that is in the same order of magnitude as a plain simulation. We describe an application of this methodology to the trip-based DRACULA microsimulation and present an illustrative example that clarifies its capabilities.
|
|
| 8. |
- Jansson, Emil
(författare)
-
Challenges with Driverless and Unattended Train Operations
- 2023
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Demand for transportation continues to increase, for both freight and passenger services. One of the most energy-efficient modes of transportation is rail. One solution to increase the attractiveness of rail transport is to introduce automatic train operation (ATO) with a high grade of automation (GoA). Driverless and unattended train operation could entail positive effects but would also bring challenges when removing the train driver. Thus, there is a need to understand the role of train drivers, especially in unplanned events. The main research objectiveis to understand the train driver roles during unplanned events and the frequency of such events. This thesis includes three papers to fulfill the research objective.This thesis studied delay logs and trackside sensor logs. A qualitative method, thematic analysis, was used to identify themes of the roles performed by train driver from the delay logs. The chi-square test statistical method was used to analyze these trackside sensor logs.Six main categories of tasks for train drivers were identified for unplanned events. Detect, Report, Inspect, Adjust, Manage passengers, and Respond to train orders. Each category was analyzed for each grade of automation by giving the responsibility for each category. The results highlight in a novel way the varied challenges between grade of automation in mainline systems. Detecting abnormalities was the most common task train drivers performed during unplanned events. Train drivers use four human senses to detect abnormalities: sight, hearing, touch, and smell. This indicates the need for onboard sensors. However, the real challenge is in processing all sensor data to gain anaccurate evaluation of any fault. One specific type of unplanned event in which the train driver is needed involves trackside sensor alarms. Freight trains are ten times more likely to trip an alarm than passenger trains. Alarms are more frequent in colder climate zones during winter months. These differences are statistically significant and indicate that not all lines and train types might be suitable for a high grade of automation.If driverless or unattended train operation will become a reality in future, many challenges must be met. This thesis gives deeper understanding of these challenges using a novel way to identify and quantify train driver tasks during unplanned events.
|
|
| 9. |
- Licciardello, Riccardo, et al.
(författare)
-
Integrating yards, network and optimisation models towards real-time rail freight yard operations
- 2020
-
Ingår i: Ingegneria Ferroviaria. - Rome, Italy : Collegio Ingegneri Ferroviari Italiani. - 0020-0956. ; 6, s. 417-440
-
Tidskriftsartikel (refereegranskat)abstract
- This paper describes the state of advancement achieved in the OptiYard research project in the use of optimisation algorithms in interaction with microsimulation of the rail-yard and surrounding network towards realtime yard management and communication with the network. Two case studies, a hump marshalling yard (mainly Single Wagon Load traffic) and a flat shunting yard (mainly intermodal traffic), were represented with state-of-the art microsimulation models, combined with innovative optimisation algorithms. Some specialistic information on the nature of the models is provided. However, the focus is oriented to railway engineers, with a description of the interactions between the models in producing outputs that are useful both to the yard dispatcher (decisions on staff, track, locomotive assignment, order of operations) and the infrastructure manager of the surrounding network (expected times of departure, availability of tracks in the yard).
|
|
