Visual Inertial Odometry for Surveillance Mobile Robot

Introduction:

Surveillance robots usually move within environments previously observed. Onboard vision sensors allow acquiring data of the environments and using that data to visit, and image more times, some places already visited. This MSc project proposal is concerned with estimating the location of a wheeled mobile robot, saving in a database some places to image again as e.g. the home location, and moving among the previously observed places.

Detailed description:

Sensing and moving in an unknown cluttered environment is based on navigation methodologies, such as visual simultaneous navigation and mapping (v-SLAM).

More in detail, is considered that the robot acquires information in a first run, and then uses that information to find previously seen landmarks and therefore repeat paths travelled in the past.

The work is organized in the following main steps:

1) data acquisition, implies moving the robot within the scene and capturing scene information (considering the Covid19 crises, this step may be replaced by use publicly available datasets)

2) data integration into a single scene representation

3) demonstration of an autonomous travelling as e.g. coming back to the home location

References:

[Cruz15] Wireless Mobile Camera, João Cruz, MSc thesis, MEEC/IST, 2015

[Pereira14] Robotic Home Sentinel, David Pereira, MSc thesis, MEEC/IST, 2014

[Mendonça14] Omni-Telepresence using Mobile Robots, José Mendonça, MSc thesis, MEEC/IST, 2014

[Ribeiro13] Nomadic Surveillance Camera, MSc thesis, MEEC/IST, 2013

Expected results:

At the end of the work the students will have enriched their knowledge in:

* Computer vision

Observations:

--

More MSc dissertation proposals on Computer and Robot Vision in:

http://omni.isr.ist.utl.pt/~jag