Autonomous vehicle surveillance: Understanding failures

QinetiQ recently collaborated with four of the University of Melbourne’s School of Engineering masters students, as part of a student/industry (Capstone) project aimed at finding a solution to the unreliability of autonomous vehicles. Alongside QinetiQ experts, the mechanical engineering and mechatronics students designed a distributed coverage algorithm suitable for a fleet of vehicles equipped with local sensing and communication capabilities.

Australia blog image

 

The project, sponsored by QinetiQ and supported by our regional communication’s expert, Michael Hall, and with academic supervision from the University’s Dr Airlie Chapman, was a year-long project across two semesters. QinetiQ’s role was to support the masters students – Chris Harrison, Yun Fang, Shuo Liu and Zhaoxi Yi – to undertake the research, while providing real-world context and interaction and issues clarification to support the development of their solution. Once a distributed coverage algorithm for a fleet of vehicles was designed, the students then simulated performance on a virtual fleet before exploring performance assurances accounting for individual vehicle failures, vehicle motion characteristics and environmental perturbations.

A step towards autonomy

Distributed sensing problems arise when multiple sensors are placed in an area and, through the amalgamation of local measurement, cooperatively survey the environment. Individual sensor failures can leave blackout areas where coverage is poor. Further, if information is relayed through the sensor network to a centralised fusion point, sensor failure can lead to a disconnectionor extended delay of the relay communication network. 

An alternative is to equip sensors to mobile platforms such as a fleet of autonomous vehicles. The problem of continual coverage of an area then becomes a fleet positioning problem. If failures in the fleet occur then the vehicle fleet should adapt, transitioning from its initial planned position to a new more performant one. It is this problem that the team are looking to solve. 

Empirical coverage

This Capstone project aimed to design a control algorithm for distributed coverage for a fleet of autonomous vehicles. One key consideration was optimising performance measures when designing the algorithm, including:

  • Time to transition between positions
  • Coverage degradation 
  • Minimising the energy to transition. 

Another significant consideration for the final energy measure was accounting for the motion of the vehicles and how the vehicles interacted with the environment, e.g. wind and terrain. 

The students have now completed their project and have presented their work at the School of Engineering’s end of year exhibition, Endeavour, and recently presented to QinetiQ’s employees at a professional development seminar.

The students presented some promising findings that will contribute to QinetiQ's overall body of work in autonomy and, with the University, QinetiQ are investigating a follow-on Capstone project.

Working with the University in this way, QinetiQ is able to offer students the opportunity of working on a real-world problem, refining their skills and knowledge in a safe environment, and interacting with industry professionals, all the while contributing to the next wave of transport innovation.

For further information on QinetiQ’s academic partnerships, please contact Alan Steele, Program Director, Engineering Centre of Excellence, QinetiQ.