The Oskillator, Artificial Force Field Highway Chauffeur
Ladda ner
Typ
Examensarbete för masterexamen
Program
Computer science – algorithms, languages and logic (MPALG), MSc
Publicerad
2020
Författare
Larsson, Oskar
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
The fully autonomous vehicle is not here yet, but a first step would be a completely
autonomous feature for a simpler subset of scenarios. In this research work, the
Oskillator is presented, a combined behavior and trajectory algorithm for a Highway
chauffeur based on artificial force fields. At each point in time an acceleration vector
is determined by generating an artificial force based on the road and the other
vehicles. The force is generated from a set of simpler force components, such as
the Lane Component for centering in lane, and the Pass Component for switching
lanes behind a slower vehicle, etc. The components are not additive as the logic for
traffic behavior is not. Instead, a composition method was developed based on min
and max operations. This way the components from different vehicles on the road
prevents the influence of the others, a single car is as much an obstruction as three.
The force is continuous w.r.t. all inputs and a damping is analytically determined
as to avoid oscillations. The host is proven to never engage in a lane switch to a
lane of another vehicle with an unsafe longitudinal distance and to stop in time
to avoid crashing upon entering the unsafe longitudinal distance behind another
vehicle. Experiments have been performed in a simulation environment to assess
the behavior of the model. It shows that the host is able to follow the road in lane
and pass when appropriate, or approach the desired headway smoothly if unable to
pass. The emergent lane switching manoeuvres have a very low lateral velocity, thus
the switch takes a long time. Several components could be extracted for standalone
use, such as Adaptive Cruise Control or Lane Keep Assistance.
Beskrivning
Ämne/nyckelord
Autonomous, Self-driving cars, Highway Chauffeur, Highway Pilot, behaviour planning, trajectory planning.