Recent work in human-robot interaction has produced experimentally validated methods for perception-aware safety assessment and intent prediction. The next step is not to expand theoretical scope, but to deploy these methods within an autonomous mobility system and validate performance under real operational conditions. This project therefore focuses on applied deployment: building a field-ready stack that detects and tracks people, dynamically prioritizes safety-critical scene elements through context-sensitive, task-driven perception, predicts pedestrian intent and motion using probabilistic, uncertainty-aware models, and uses those outputs to support proactive rather than purely reactive navigation behavior in shared spaces.
The work will be executed in a controlled and measurable progression from environment baseline creation, to system mapping, to module integration, to pilot operation, generating operational data and insights that can inform further research and evaluation. The result will be a deployable prototype validated on a real KACST environment with explicit acceptance criteria, quantitative safety metrics, and operational evidence suitable for internal evaluation and potential external reporting.
The objective of this project is to deploy, adapt, and validate human-aware perception, safety, and intent prediction methods on an autonomous mobility platform operating in a shared human environment, and to evaluate their performance using real-world operational metrics, demonstrating the system’s capacity for proactive, socially-aware interaction and its readiness as a foundation for broader collaborative intelligent system development.