Examenspresentation av Anton Gustavsson

Examiner: Emmanuel Dean

Abstract:
This thesis presents the development and evaluation of two methods for detecting
and tracking objects from radar data, for a fleet of autonomous transport robots
operating in a industrial environment. The system is developed as a safety feature
to avoid collision with any object or coworker in the robot’s direction of travel.
A 60 GHz pulsed coherent radar sensor is used to detect radial distance to any
objects within the scene. The evaluation includes using different configurations,
physical installations and methods for speed estimation with the sensor in order
to find a satisfactory performance. Two methods were developed and tested: One
method based on using pulse-timing and the Doppler effect of the returned signal
to determine the object’s radial distance and relative velocity. The other method
uses triangulation and Kalman filtering to track objects between measurements and
calculates the velocity from the estimated data. During the evaluation, scenarios
with known data were acted out and subsequently compared with the data, gener-
ated from the sensor. The main ability of interest is to consistently detect objects.
As well as accuracy in angle, relative position, relative velocity.
From the results it is concluded that the method using range-Doppler is the most
reliable. To estimate speeds, using the Doppler effect proved most accurate as well.
Also, the performance of both systems are greatly dependent on the configuration
of the radar sensor. Significant efforts had to be invested into finding a satisfactory
parameter set up. Tests were performed separately from the robot and thus perfor-
mance cannot be guaranteed in an industrial implementation, even though results
show promise.

Welcome!

Anton and Emmanuel