1. the complete title of one (or more) paper(s) published in the open literature describing the work that the author claims describes a human-competitive result; Expert Competitive Traffic Light Optimization with Evolutionary Algorithms 2. the name, complete physical mailing address, e-mail address, and phone number of EACH author of EACH paper(s); Yann Semet Thales Research and Technology 1 Avenue Augustin Fresnel 91120 Palaiseau, France +33 1 69 41 55 16 ysemet@gmail.com Benoît Berthelot CDVIA 2 rue Suchet 94700 Maisons-Alfort, France +33 1 43 53 69 47 b.berthelot@cdvia.fr Thierry Glais Thales Revenue Collection Systems Rue de la Mare aux Joncs 91220 Le Plessis-Paté +33 (0) 1 69 88 56 46 thierry.glais@thalesgroup.com Christian Isbérie CDVIAjjjjjj 2 rue Suchet 94700 Maisons-Alfort, France +33 1 43 53 69 47 b.berthelot@cdvia.fr Aurélien Varest CDVIA 2 rue Suchet 94700 Maisons-Alfort, France +33 1 43 53 69 47 a.varest@cdvia.fr 3. the name of the corresponding author (i.e., the author to whom notices will be sent concerning the competition); Yann Semet ysemet@gmail.com 4. the abstract of the paper(s); We present a complete system to optimize traffic lights green phases and temporal offsets based on a combination of microscopic simulation and black box, evolutionary algorithms. We also report the outcome of an AI versus experts comparison workshop conducted with our algorithm and seasoned experts from a specialized traffic engineering office. Experimental results indicate that the proposed algorithmic scheme significantly outperforms expert efforts. Our system entails a memetic (genetic+gradient) calibration module to adapt the Origin/Destination (O/D) matrix to current traffic conditions, an inoculation procedure to incorporate existing traffic light programs, genetic multi-objective optimization capabilities and sound metrics. Experiments are conducted over several real world datasets of operational sizes from the Paris outskirts and various other French urban areas. Our experimental outcome is threefold. First, we report the success of the memetic calibration module in adjusting the simulators O/D matrix to a point with variation levels corresponding to recorded sensor data. Second, we confirm the ability of the system to obtain significant gains on that sound basis: gains ranging from 15% to 35% are consistently reached on both traffic jams reduction and pollutant emissions. Most importantly, we report the outcome of the comparison workshop: a formalized methodology followed by experts to manually optimize traffic lights, iterative experimental logs tracing the application of that methodology to two real world cases and comparable results obtained by the algorithm on the same cases. Results indicate that the AI module performs significantly better than experts in both speed and final solution quality. 5. a list containing one or more of the eight letters (A, B, C, D, E, F, G, or H) that correspond to the criteria (see above) that the author claims that the work satisfies; A, D, G, H 6. a statement stating why the result satisfies the criteria that the contestant claims (see examples of statements of human-competitiveness as a guide to aid in constructing this part of the submission); (A) The result was patented as an invention in the past, is an improvement over a patented invention, or would qualify today as a patentable new invention. The system is currently undergoing two specific, distinct patent application processes to procure protection for the idea of using an algorithm to automatically produce non trivial traffic data results needed in solving difficult regulation problems that pose serious difficulties to traffic engineers on a daily basis: (1) producing optimal decision trees to map traffic sensor data to dynamic traffic lights based actions and (2) producing classifiers to correctly identify traffic conditions. The results produced in both cases (decision trees and classifiers respectively) have value by themselves and are all that matter to traffic engineers, the patent applications are not specific as to how they are produced although experimental evidence is proposed to substantiate the claim using Evolution Strategies and a combination of Machine Learning algorithms respectively. The two applications are the following: Title: Decision Tree Optimization for Urban Traffic Regulation Targeted Countries: France and thre rest of Europe Status: Validated by internal Thales Intellectual Property Protection jury, French patent office filing targeted for summer 2019 Title: Algorithmic Processing Chain for Urban Traffic Macro-Regulation Targeted Countries: France and the rest of Europe Status: Validated by internal Thales Intellectual Property Protection jury, French patent office filing targeted for summer 2019 (D) The result is publishable in its own right as a new scientific result independent of the fact that the result was mechanically created. Our work was a published in the proceedings of a peer-reviewed traffic science conference devoted to advances in vehicle technology and transport systems with no direct link to algorithms or artificial intelligence. We believe this gives credit to our work from an application only perspective and confirms the interest of the corresponding automatically produced results on their own merit. We think it also sanctions the relevance, importance and difficulty of the problem at hand for this particular application community. We also consider it particularly important that the traffic community, as an independent body, availed and recognised the validity of the competition we organized between algorithms and experts (see criterion H). Additionally, our paper contains a section devoted to reverse engineering with a specific example of how valuable, pragmatic lessons can be drawn from the automatically produced result itself and opposed to the traced expert methodology we report in another section in order to adjust it for future endeavors. Following the conference, our paper was short-listed for inclusion, as a 25 pages extended version, in a Springer book to be published this summer. (G) The result solves a problem of indisputable difficulty in its field. Traffic regulation is a pervasive conundrum for all major cities, and an increasingly worrying public health concern as vehicle emitted pollutants (CO2, NOx and particulate matter among others) are an officially recognized source of pathologies such as cancer, birth defects and respiratory conditions. Officials and engineers therefore strive to achieve better regulation and to that end, traffic lights are a convenient lever with very significant impact. No new infrastructure needs to be built and optimization effects are substantial and immediate. By smartly orchestrating flows, one can indeed manage to act on both vehicle throughput (less traffic jams) and traffic structure (potentially less pollution). The difficulty is that the underlying mathematical problem, an optimization one, is actually extremely difficult. It is very simple to state and explain but falls prey to several forms of severe mathematical difficulty. It is, particularly, highly multi-modal, strongly non linear, disturbingly chaotic, very eptistatic and highly non separable. These difficulties have immediate, maddening, consequences on practitioners who are responsible for solving it: they almost cannot perform the smallest of changes on one variable without facing unpredictable and drastic consequences on the rest of the problem. This is in large part due to the intricate nature of the road network which is indeed a dense graph of highly connected intersections. There is also an intrinsically dynamic quality to the traffic phenomenon that makes it difficult to characterize, predict and master. As a result, preparatory or reoptimization traffic light studies are extremely tedious, long and costly. They require trained engineers, usually experts with specialized graduate engineering degrees and ample field experience completed with local knowledge. These studies can last for months and often end up with heuristic proposals with questionable statistical validation. Traffic control is, consequently, a very significant problem and line of effort for the corresponding engineering and scientific community, which devotes to it books, scientific articles and entire conference tracks, usually under the Smart City banner (see for example the programme of the 2019 World Congress on Intelligent Transport Systems). Our system efficiently solves the traffic light problem with evolutionary search. It can work against traffic jams only or in multi-objective mode, trying to minimize pollutant emissions as well. It was tried against several real-world problems, carefully calibrated on sensor data and produced statistically validated results. Significant quantitative gains were obtained with respect to currently deployed traffic light plans, ranging from 15% to 35% depending on metric and problem instance. (H) The result holds its own or wins a regulated competition involving human contestants (in the form of either live human players or human-written computer programs). In order to assess the ability of our system to save time and increase the efficiency and agility of traffic studies usually performed by experts, we chose to organize, on top of our in-lab experimental campaign, a competitive workshop between our algorithm and 3 real experts with a cumulated 51 years of experience, from an independent, specialized traffic engineering office (CDVIA): C. Isbérie, M.Eng., B. Berthelot, M. Eng. and A. Varest, M. Eng., M. Sc. Several sessions were conducted on real world problems. In order to maximize the validity of the comparison, we decided to: (1) organize a fair, closed doors competition without high performance computing for the algorithm and on the exact same problem instances and metrics (2) procure all necessary comfort to the experts so they can produce maximum possible gains: unlimited time and trials, comfort of their own office, all (non-AI) tools allowed, etc. (3) ask the experts to write down, during a separate preliminary workshop, their methodology explicitly in details and trace all corresponding followed steps during the competition (4) write down the exact composition of our algorithm as well and trace runs (5) aggregate and report expert methodology, logs, algorithms and competition rules and results in a single comprehensive paper and apply for publication in a traffic science conference for it to act as an independent validation body for both the comparison methodology and the value of the produced results. Results of the competition on two real-world test cases show that the evolutionary algorithm clearly outperforms experts in both speed and final solution quality. On the larger test zone, with 12 signalized intersections located on a major traffic axis in the immediate northern outskirts of Paris, experts took 3 hours and 45 minutes to produce a final optimization gain of around 7%. The evolutionary algorithm, run simultaneously on a standard laptop, managed to reach 35% in a little under 2 hours and crossed the level of gain obtained by experts in under 6 minutes. 7. a full citation of the paper (that is, author names; publication date; name of journal, conference, technical report, thesis, book, or book chapter; name of editors, if applicable, of the journal or edited book; publisher name; publisher city; page numbers, if applicable); Semet Y., Berthelot B., Glais T., Isbérie C. and Varest A. Expert Competitive Traffic Light Optimization with Evolutionary Algorithms, Proceedings of the 5th International Conference on Vehicle Technology and Intelligent Transport Systems, Heraklion, Greece, May 3-5 2019, 199-210. 8. a statement either that "any prize money, if any, is to be divided equally among the co-authors" OR a specific percentage breakdown as to how the prize money, if any, is to be divided among the co-authors; Prize money, if any, is to be divided equally among the co-authors. 9. a statement stating why the authors expect that their entry would be the "best" We argue that our entry's competitive strength is twofold. First, we efficiently address the significant societal problem of regulating traffic and its consequences on public health and sustainable concerns: better traffic means more time, less stress, cleaner air and better health for all citizens. It also augurs of considerable savings and improved productivity for both public traffic regulation budgets and the private traffic engineering sector. Our work shows that an automated system can procure human-competitive help in that worthy endeavor by leveraging the power of evolutionary search in solving an indeed very difficult and intricate mathematical problem. We believe our entry also has particular merit in that it offers a comprehensive answer to the problem with experimental evidence for all the various subcomponents of the general problem (simulator calibration, traffic condition identification, static control optimization and dynamic control optimization). More importantly, we expect our entry to be particularly interesting because it puts a strong focus on the expert-competitive character of the results we obtain. Through the competition we organized, we tried to obtain explicit, traceable and hopefully convincing evidence that an evolutionary algorithm can indeed do better than actual experts on this particular problem and in the exact same conditions. The results we obtained were published in the traffic science community and are clear-cut: on a real-world problem instance, the evolutionary algorithm can do just as well as hours worth of expert effort in a handful of minutes and improves final gain figures by a factor of five. 10. An indication of the general type of genetic or evolutionary computation used, such as GA (genetic algorithms), GP (genetic programming), ES (evolution strategies), EP (evolutionary programming), LCS (learning classifier systems), GE (grammatical evolution), GEP (gene expression programming), DE (differential evolution), etc. MOEA (NSGA-II variant) and Evolution Strategies. 11. The date of publication of each paper. If the date of publication is not on or before the deadline for submission, but instead, the paper has been unconditionally accepted for publication and is “in press†by the deadline for this competition, the entry must include a copy of the documentation establishing that the paper meets the "in press" requirement. May 2019.