Workshop on Radar and Computer Vision Fusion for Failsafe Design in UAVs (p...

UPDATE: Due to the high demand for participation in the workshop, the event will be hosted in Room 5, at 1st Floor of UNIWA's Conference centre in Elaionas Camp, while availability of seats has been extended to accommodate demand.

Presented by: Dr Simon Winberg, University of Cape Town, South Africa

Organizers:

- Computer Networks and Services Lab (CONSERT), Department of Electrical and Electronic Engineering, University of West Attica, GREECE

- Department of Electrical Engineering, University of Cape Town, SOUTH AFRICA

The Workshop is co-funded by the Erasmus+ ICM Programme, KA107-80664 (Greece - S. Africa, 2019-2020).

Failsafe design is based on the principle of understanding the limitations of a system being built and – as much as engineers may dislike admitting it – that there is potential for parts of that system to fail or not to always work reliably. Redundancy is one commonly used approach of implementing failsafe designs, whereby a primary system is backed-up by a secondary system.
Redundancy can be on the level of an entire system, on individual sub-systems, or even smaller parts. This workshop concerns a radar and computer-vision sensor fusion system, whereby sensor fusion provides richer navigation data for the situational awareness and control of an Unmanned Arial Vehicle (UAV). This allows the UAV achieve mission objectives more effectively than with only the one or the other sensor feed. This approach also gives the added benefit of providing partial redundancy, whereby the vision sensor can stand-in for the radar or the radar can stand-in for the visual sensors. The vision component in this case comprises dual cameras, providing a greater field-of-view as well as utilizing stereopsis to obtain distance information. But fusing relatively slow stereopsis depth data with very fast radar range and Doppler data does not have an obvious straightforward solution, especially when considering the application concerned.
In this workshop, we will encounter various problems and explore solution strategies for this application. The context of operation is around support for urban terrain applications, e.g. targeted environmental monitoring (air and water pollution). Urban terrain exhibits a range of challenging contextual characteristics, for example, there is often a built environment full of sudden and possibly indistinct obstructions, corners and air corridors that need sufficient local situational awareness data together with rapid control to make the operation of the UAV safe as well as secure. In these situations, neither a radar on its own, nor a single visual camera, would be sufficient for the reliable operation of UAVs. Nonetheless, in the theme of failsafe operation, if either of these UAV sensors becomes compromised, it is still preferably for the UAV to maintain some form of operation so that it can be extracted from the operation zone.
The aim of this workshop is to both present participants with insights into the technologies concerned, to reflect on some lessons that were learned the hard – and expensive – way, and to present a sequence of brainstorming and solution-seeking activating to give participants a start on engaging with these issues and to deepen their awareness of challenges and solutions for this research area. This workshop has been designed for a 2h duration; due to this limited time we will not have opportunity to delve into detailed design and coding solutions; this workshop is thus more design around activities which do not require participants to have access to computers.