Authors:
J. C. F. Allaire
1
;
J. M. P. Langlois
2
;
G. Labonté
1
and
M. Tarbouchi
1
Affiliations:
1
Royal Military College of Canada, Canada
;
2
Polytechnique Montréal, Canada
Keyword(s):
World representation, Real-time, Evaluation, FPGA, UAV, Path planning.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Computational Intelligence
;
Evolutionary Computing
;
Representation Techniques
;
Soft Computing
Abstract:
Unmanned aerial vehicles (UAV) are subject to unforeseen events in harsh environment. Embedded autonomous real-time path re-planning is a possible solution to this issue. Evolutionary algorithms have shown to be an excellent means to optimise the generation of UAV paths but their slow iterative process prevent them to be used for real-time computation. Part of that challenge resides in the computational demanding task of path feasibility evaluation, where each single segment of the generated path needs to be certified ‘collision free’. State of the art algorithms require computationally demanding pre-processing of the world representation, which is too time-consuming for real-time computation. Taking advantage of advancements in the Field Programmable Gate Array (FPGA) technology, this work has evaluated a new feasibility evaluation technique that analyses the path directly from the raw data of the world representation, using two levels of resolution: a high resolution map used close
to the UAV, and a low resolution map used far from the UAV. This technique has been implemented on an FPGA and tested in simulation. Timing results (more than 500 map cells evaluated within 5 μs) demonstrate that the two-tiered resolution technique opens up avenues to real-time UAV path re-planning using evolutionary algorithms.
(More)