Waqas Ali

I am a PhD student at the Division of Robotics, Perception, and Learning at KTH Royal Institute of Technology. My research focuses on building resilient perception systems for autonomous robots. By developing life-long mapping capabilities for unstructured, complex environments using consumer-grade LiDARs, my research aims to operate with minimal human intervention and directly support downstream tasks.

Core Research Areas & Contributions:

Robust Large-Scale SLAM: Improving globally consistent mapping in GPS-denied environments using HD-map priors and developing highly generalizable LiDAR place recognition pipelines resilient to domain shifts.

Lifelong Map Maintenance: Creating probabilistic persistency models to dynamically refine 3D maps and intelligently filter out dynamic or quasi-static objects as environments evolve.

Probabilistic 3D Scene Graphs: Building hierarchical, semantically rich spatial representations that capture fine-grained geometry without sacrificing scalability, while explicitly modeling sensor noise and perceptual uncertainty.

High-Level Reasoning & Open-Set Perception: Transitioning to task-driven dynamic vocabularies, and integrating probabilistic scene graphs with vision-language foundation models to ensure reliable robotic reasoning.

Publications:

[1] Ali, W., Cai, Y., Jensfelt, P., & Nguyen, T. M. (2026). ProbPer-LiLo: Probabilistic Persistency Modeling for Life-Long Mapping. IEEE Robotics and Automation Letters, 11(3), 2530-2537.

[2] W. Ali, P. Jensfelt and T. -M. Nguyen, "HD-Maps as Prior Information for Globally Consistent Mapping in GPS-Denied Environments," 2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), Edmonton, AB, Canada, 2024.