With the aid of sensors such as cameras, heat cameras and radars mounted on aerial vehicles, items on the ground can be found more rapidly, collisions with other flying objects can be avoided, missing people can be found, and forest fires can be discovered and monitored. There is no lack of important applications for this type of sensor platform.
“Today, even manned aeroplanes are flying nearly autonomously: the pilot is instead fully occupied with controlling all the sensors. If we can automate the control, the pilot will gain more time to interpret the information that the sensors provide”, says Per Boström-Rost, industry-based doctoral student at the Division of Automatic Control at LiU and employed by Saab Aeronautics.
Informative path planning
His licentiate thesis describes studies of what he calls “informative path planning”. This term describes a method in which systems use the obtained sensor measurements to influence how the vehicle subsequently moves.Per Boström-Rost. Photo credit Magnus Johansson“Just as we do, the plane may get the impression from a sensor signal that there’s something interesting over there, and then guide the sensor platform in that direction”, he explains.
The idea is that the sensors will be able to track a moving object, on the ground or in the air, independently of how it moves. In this case, the sensor system is to be controlled such that the future measurements are as useful as possible, and decisions must be taken based on assumptions about which measurements will be needed. It’s not known in advance how the tracked object will move, which means that the autonomous planning of how the sensor platform is to move takes place with considerable uncertainty. The problem can be solved in a planning algorithm by making several assumptions.
Without being detected
Per Boström-Rost describes in the thesis not only how a path can be planned when the sensors have a limited field of view, but also how the vehicle can be caused to monitor an object or location without itself being detected.Per Boström-Rost and Kristoffer Bergman Photo credit David Brohede“We sometimes want to collect information without being seen. This is another path – describing how the sensor platform should move in order to collect as much information as possible without being detected itself – that can be planned using optimisation algorithms”, he says.
Per Boström-Rost is one of approximately 58 industry-based doctoral students who are working within the huge Wallenberg AI Autonomous Systems and Software Program (WASP). A total of approximately 185 doctoral students are currently working within the programme, and more are expected.
On Informative Path Planning for Tracking and Surveillance
Boström-Rost, Per, Department of Electrical Engineering, Division for Automatic Control, Linköping University, Faculty of Science and Engineering, 2019. Supervisors Gustaf Hendeby and Daniel Axehill.
Translated by George Farrants