Localisation

Our machine learning-based robotic perception for accurate localisation, semantic mapping of the lunar terrain, and object detection for close-range manoeuvring between rovers
In ICRA 2022. (Won 3rd place in NASA Space Robotics Challenge and an Innovation award).

Achieve a lightweight & scalable solution for place recognition
IEEE Robotics and Automation Letters 2021.

Repeatability of an interest point is not constant but a function of time
In ICRA 2021

Propose 2 filtering approaches for visual localization: Monte Carlo Visual Localization, and Hidden Markov Model.
Neural Computing and Applications 2020 (Special Issue on Best of DICTA 2019).

Explore the sparsity of topological map to improve the inference time for place recognition
In ICRA 2020

Propose Monte Carlo Visual Localization, which ensures the temporal consistency of localization inference.
In DICTA 2019.

Highlight the need for continuous data collection in achieving a robust place recognition system. This drives our proposal for a scalable solution based on the Hidden Markov Model.
In ICCV 2019

We propose an efficient outlier-robust algorithm to tackle INS and GPS sensor fusion problem for pose estimation.
In IROS 2019

Develop an on-device system for large-scale visual localisation
IEEE Transactions on Image Processing 2018.

Present a software to collect simulated data from computer game GTA V. Users have fully control of camera poses and environmental conditions.