This research program addresses the fundamental problem of how a robot or autonomous vehicle uses perception to create maps and calculate and track its location in the world.
Research questions include addressing how the appearance of a place changes in relation as a function of time, season, weather, viewpoint and environment type; how understanding context and semantics can enhance performance; how lifelong reliability can be achieved as the world continually changes; the relationship to neurological structures and behavioural mechanisms used in animal and human navigation; and how new perception technologies can be applied to this problem.
This program will build on already considerable industry collaborations in high value sectors including mining, defence and look to expand into logistics, construction, space and remote operations. An extensive program of societal engagement will continue built on a large number (50+ per annum) of public or sector-based talks, workshops, media, and large-scale engagement activities with the public.
- ACRV Picking Benchmark
- Amazon Picking Challenge (2016)
- Reliability in Deep Machine Learning and Uncertainty for Object Detection
- Scene Understanding and Semantic SLAM
- Reinforcement Learning for Robot Navigation
- Learning Robotic Navigation and Interaction from Object-based Semantic Maps