PTV releases ModelGPT – a new AI-powered system to build trans-portation models within seconds
Karlsruhe 1 April 2023. PTV Group, the market leader for mobility software, launches a new revolutionary mobility system based on artificial intelligence. With ModelGPT, users can get their transportation model for any city or region worldwide and any desired use case within just a few seconds. ModelGPT is the first application of a new Model AI product family, with further application to be launched in 2023.
ModelGPT is based on a deep neural network which was trained on a large dataset of historical traffic data generated out of the extensive experience in modelling projects by PTV and its user community. This allows the Generative Pre-trained Tranformer (GPT) to learn complex patterns and relationships between different variables. Users can simply type in their requirements, parameters, and scenarios to be considered in text form from which ModelGPT then generates the transportation model within seconds. An additional layer of training – Reinforcement Learning with Human Feedback (RLHF) – enables the AI-powered system not only to understand and implement the user’s requests, but also to interact and ask questions in case of ambiguities.
After a short processing time of only a few seconds the transportation model can be immediately used for analysis and visualization for several use cases.
Peridot Lavender Head of AI Modeling at PTV said: “ModelGPT has the potential to revolutionize the way we shape our mobility ecosystems, as AI gives impressive capabilities to modelers. At the same time, we know how valuable the human resource is to build powerful models. Therefore, our tools using AI empower human planners to make modelling even easier and inclusive, and to completely focus on strategy and actions to shape a safe and efficient mobility that benefits all of humanity.”
ModelGPT was trained on massive amounts of data. It has 235 billion parameters and was trained on 725 gigabytes of existing modelling and simulation examples. Since the system is still in its research and learning preview phase, planners can now use it for free to further train and improve the system
In addition to ModelGPT, PTV Group is developing several additional applications of the new Model AI product family. Already in planning are ModelLensa, a tool to turn models into art, ModelGram, a program to share model design with others, MoDELL-E, an AI-System that creates digital twin solutions as realistic as Matrix and MetaModel, which empower users to enter models in the metaverse.
ModelGPT can be used for free at: https://company.ptvgroup.com/en/modelai
PTV. Empowering Mobility as part of Umovity.
PTV Group is a leading global software company for traffic planning, simulation, and real-time management. With more than 40 years of experience in the fields of mobility, the Germany-based company provides software products based on proprietary algorithms ranging from microscopic and macroscopic modeling and simulation of traffic to real-time traffic management, benefiting more than 2,500 cities and municipalities. Bridgepoint acquired a majority stake in the company in January 2022 in order to further accelerate its growth together with shareholder Porsche Automobil Holding SE. Since 2023 PTV Group and Econolite are united under the brand Umovity.
Umovity (PTV Group & Econolite). Mobility for Humanity.
Umovity is a global market leader for end-to-end traffic management and transportation technology that takes a holistic approach to provide safer, smarter, and more sustainable Mobility for Humanity. Uniting the best-in-class solutions for intelligent traffic management systems, cloud-based adaptive traffic control, real-time traffic management software, controllers, cabinets, and sensor products by Econolite, and simulation and predictive modelling software by PTV, Umovity enables innovators and decision makers in politics, municipalities, and industry to shape smart and livable communities, multimodal mobility, as well as Connected and Autonomous Vehicles (CAV). For more information, visit www.umovity.com .