Complex Modern Agent Behaviours
Our group investigates techniques to diversify and modernize agent behaviours. Here is a screenshot from a simplified transit scenario where agents use event-driven and parametric behaviour trees in conjunction with probabilistic visual attention fields to model egocentric distractions like the use and impact of cell phones on crowd movement in modern urban environments.
High Fidelity Physical Crowds
The future of synthetic crowds for simulation, animation, safety-critical analysis, and design is ultra-high fidelity physically enabled crowds. Our group explores and developed new techniques in Reinforcement Learning, Multi-agent Reinforcement Learning, Deep Learning, Steering, Biomechanics, and Physical Character Control to support the next generation of crowds models and applications. These approaches support the modelling and animation of more diverse and representative crowds with a long-term goal of modelling disabilities, mobility devices, and cognition in synthetic crowds.
Human-Augmented Intelligence in Computer Aided Design
We improve design pipelines and strengthen outcomes and deliverables by augmenting designer skills with machine learning, artificial intelligence, and optimization via static and dynamic analysis. Our augmented pipelines produce environments that are human-aware and predictive to a broad range of scenarios.
Game Analysis, Optimization, and Adaptive Procedural Generation
Building more engaging games by understanding how players move and engage in the environment. Analyze and model the underlying difficulty to produce target level difficulties through optimization. Use these levels to procedurally augment player experience or provide options for player controlled experiences
We develop game platforms based on high-fidelity clinical tracking and measurement tools. Using these platforms we engage in participatory design and development of visual metaphors and game mechanics for reaching rehabilitation goals. We drive characters and visualizations using computational geometry to capture clinical motor speech targets. This approach augments clinical practice to improve attrition rates in a variety of rehabilitation and therapy settings.
Augmenting Design Communication with Virtual Reality
Often complex designs are difficult to engage with. We explore and build interactive technologies using augmented and virtual reality to facilitate and streamline stakeholder and collaborator communication. Our platforms have highlighted and verified design flaws and errors while improving the accuracy of the perception of design choices.
We acknowledge with respect the Lekwungen peoples on whose traditional territory the university stands and the Songhees, Esquimalt and W̱SÁNEĆ peoples whose historical relationships with the land continue to this day.
The GAIDG Lab is interested in broad ways of knowing and understanding and recognizes that ideas, research, and the people involved in these may come from many different places and backgrounds. We also recognize that historically, and very much presently, work in the fields in and around Computer Science and particularly Machine Learning and Artificial Intelligence has had a role in and facilitated colonialism, racism, ableism, transphobia, and homophobia. Additionally, the computational resources required for modern AI, ML, and data mining have a direct impact on the environment. The GAIDG Lab is committed to addressing and reflecting on these issues in our work, hiring and training practices, our community output, and our outreach.
Department of Computer Science
Room 648, Engineering and Computer Science (ECS) building
University of Victoria