Mining For Meanings In Robot Maps

Title Mining For Meanings In Robot Maps
Summary To build a hybrid map by augmenting the intrinsic kinematic model of a mobile robot to a spatial map, and semi-supervised learning of meanings towards self/situation awareness.
Keywords robotics, mapping, semantic maps, unsupervised semantic mapping, data mining, kinematic model, situation-awareness.
TimeFrame Spring 2017
References Pronobis, Andrzej, and Rajesh PN Rao. "Learning Deep Generative Spatial Models for Mobile Robots." arXiv preprint arXiv:1610.02627 (2016).

Khalil, Wisama, and Etienne Dombre. Modeling, identification and control of robots. Butterworth-Heinemann, 2004.

Shahbandi, Saeed Gholami, Björn Åstrand, and Roland Philippsen. "Semi-supervised semantic labeling of adaptive cell decomposition maps in well-structured environments." Mobile Robots (ECMR), 2015 European Conference on. IEEE, 2015.

Prerequisites Programming (preferably C++ or Python), Machine Learning, Data Mining. Bonus: Mobile Robots (kinematic/dynamic modeling), ROS.
Supervisor Saeed Gholami Shahbandi, Björn Åstrand
Level Master
Status Open

Each region of a spatial robot map potentially has a meaning (semantics). For instance the map of house could be segmented into kitchen, corridor, bedroom, etc. The fact that these meanings are generated from what humans understand of their surrounding, is crucial for a successful communication (e.g. in HRI). On the other hand a robot becomes more “situation-aware” by knowing the semantic of its surrounding. The aim of this project is to integrate the kinematic/dynamic model of the robot into the spatial map of the environment. And employ an unsupervised method to identify the semantics of the environment while the “self” of the robot is also reflected in the spatial map.
The expectation is to bridge a robot’s self-awareness (e.g. traversability of a path based on its intrinsic models and the terrain), to the situation-awareness that is supposedly capable of estimating future state of the situation.
Research Questions
How to integrate the intrinsic kinematic/dynamic models of a mobile robot seamlessly into the spatial map of the environment? How to identify ego-centric meanings that emerge from the integration of spatial maps and robot’s kinematic/dynamic model?
Simulation, and experimental results the lab environment.