Abstract
Integrating generative artificial intelligence (AI) with interactive digital canvases, such as geospatial maps, can introduce high latency and computational costs, particularly during continuous user interactions like panning or zooming. The disclosed technology describes a system for kinematic-based predictive filtering. This system can monitor the kinematic properties of a user's navigational gesture, such as deceleration, to predict a potential final viewport before the interaction concludes. This prediction may allow a server to proactively cull a large master dataset into a smaller, contextually relevant subset for processing by a generative AI model. By performing data culling and AI inference in parallel with the user's gesture, the system may deliver a generated response with reduced latency, potentially improving the responsiveness of the user experience when interacting with large datasets.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Shanmuga Sundaram, Prashanth, "Kinematic-Based Predictive Filtering for Low-Latency Generative AI", Technical Disclosure Commons, (October 12, 2026)
https://www.tdcommons.org/dpubs_series/12093