Persistent Non Player Characters (NPCs) in many modern video games follow schedules guiding their routines and behaviors over time as the player engages in play inside the game’s virtual world. In a game like Ubisoft’s recently released Watchdogs: Legion, where schedules are a player-facing game mechanic, a robust scheduling system is highly important and critical to the game’s financial and critical success.
Making good schedules for NPCs in this context, however, is surprisingly complex.
Ubisoft records the interaction between its customers and its servers in large execution logs (also called traces). Any failure of the system is thus recorded therein. However, the considerable size of these logs considerably hinders their effective use by analysts and developers. We propose an automated method to detect failing executions, and furthermore to characterize the features that are common to clusters of failing instances. The approach will be based an machine learning algorithms, and will produce clusters of failing traces with common features.
The goal of this research project is to develop novel technics to teach artificial agents how to play complex video games using reinforcement learning and demonstrations. Namely, we wish to propose a novel approach for learning from demonstrations, in which an agent simultaneously learns a behavior and the corresponding reward signal. This training procedure will rely on generative adversarial imitation learning in order to learn from expert demonstrations (in our case from players).
The goal of this research project is to develop novel techniques to solve different tasks for character animation using deep neural networks and generative modeling. Namely, we wish propose a novel approach for transitions generation, in which clips of character animation can be linked together with a novel clip. This transition will be generated by a specifically designed recurrent neural network that should make use of recent advances in adversarial learning in order to produce realistic animations.
Ubisoft has an extensive database of 3D scanned heads. It would be convenient to use it to mix-and-match parts of characters to create new human-like character heads. Lets say we want to adjust medium-scale features of the face, such as replacing the nose of one character with another nose. We will design an editing workflow allowing the artist to create a new nose from mixtures of noses found in the database.
The goal of this project is to explore the use of log analytics and machine/deep learning techniques to improve Ubisoft operational intelligence. Logs contain a wealth of information, but often hindered by the lack of best practices, tools, and processes. Despite the importance of logging, the area has not evolved much over the years. At Ubisoft, logs are used extensively for various system diagnosis tasks. The analysis of logs, however, is usually performed manually, limiting the full potential of the information contained in logs.
Ubisoft has an extensive database of characters heads represented as polygonal meshes. Those come from two primary sources: processed 3D scans and models made by artists. It would be convenient to use this database to mix-and-match parts of characters to create new human-like character heads. Lets say we wish to replace the nose of one character with another nose. We want the junction between the nose and the surrounding areas to be as seamless as possible while accommodating for the new nose, which could have a different size.
Join a thriving innovation ecosystem. Subscribe now