15-10-2012, 01:38 PM
Development of a Collision Avoidance Algorithm Using Elastic Band Theory
Development of a Collision.pdf (Size: 933.78 KB / Downloads: 33)
Abstract:
This paper presents a new Collision Avoidance (CA) Algorithm which uses Elastic Band
Theory. Researchers tried to develop warning systems to avoid collisions which warn drivers of possible
collision risk with audio and or visual signs. However, these systems are not sufficient for avoidance of a
collision in situations where the driver gives no response to the warnings. CA System is a kind of Active
Safety System which takes control of the vehicle for a couple of seconds and applies emergency maneuver
when the collision is unavoidable through driver action alone. The proposed CA algorithm uses Elastic
Band Theory which is an obstacle avoidance method used in robotics. This paper aims to introduce this
theory applied with modifications to road vehicle based systems and presents realistic simulation results
using high fidelity vehicle models with several different collision scenarios.
INTRODUCTION
Traffic crashes cause deaths of millions of people every year.
Researchers tried to overcome this fatal problem by
developing passive safety systems like seat belts, air bags and
crash zones. Although these passive measures helped a lot,
there must be more effective ways of holding accidents at
acceptable levels. This goal is closer to realization through
advances in preventive and active safety systems (Ararat and
Güvenç, 2005). Collision Warning (CW) System is an
important jump from passive to active safety systems. CW
System tries to detect any collision risk between two vehicles
by means of radar and internal vehicular sensor information.
If the system detects collision risk, it will warn the driver so
that a possible accident can be avoided (Ararat et al., 2006).
Although a CW System is an efficient way of avoiding
possible accidents, these systems become useless when the
driver gives no response to the warnings. Collision
Avoidance (CA) System takes action and applies emergency
maneuver in these kinds of situations.
ELASTIC BAND THEORY
Elastic Band is a kind of obstacle avoidance method used in
robotics which was first proposed by Quinlan and Khatib
(1993). A deformable predefined path is modified by internal
and external forces acting on the band. Internal forces are like
spring forces which hold the band together while external
forces are like artificial potential forces which keep the band
away from obstacles. Figure 1 presents the schematic
representation of the elastic band with internal and external
forces acting on the band.
COLLISION AVOIDANCE ALGORITHM
Collision Avoidance Algorithm (CAA) is composed of main
algorithm, error detection subsystem and the decision
algorithm. Algorithm uses laser, digital map and driver as an
input source. Laser data gives obstacle data where digital
map satisfies the predefined path. Algorithm uses driver data
in decision part. Collision Detection Algorithm (CDA) which
is introduced in ([2]) is a kind of kinematical based algorithm
that provides Time to Collision (TTC) information.
SIMULATION STUDY
Since simulations should be done in an environment close to
reality with advanced vehicle models, Carsim 6.05 is used for
testing purposes. Carsim provides connection with Simulink
to realize testing of control algorithms developed in
Simulink. Conventional controllers are used at the first step
for trajectory tracking. Three different configurations with
different scenarios are tested with developed algorithm.
Following subsections give detailed results of the algorithm
for these configurations.
CONCLUSION
Elastic Band is a strong method which combines the
simplicity of the potential field method with the accuracy of
the trajectory generation based methods. This strong method
was modified for vehicle based applications. Reactive forces
that push the vehicle away from the other objects in an
environment was first modified. Uniformity of the band was
satisfied by changing the method for the situation which has
objects in the vicinity of the band. Alternative paths were
developed considering the restrictions of the vehicle
dynamics. Error detection subsystems were developed for
searching available modified alternative paths. Proposed
algorithm was tested with Carsim 6.05 for different scenarios
and different configurations. Algorithm gave acceptable
results for Rear-End, Lane Change and Intersection Point
Collisions. Although conventional controllers gave
acceptable results, performance of the algorithm might be
increased with much more advanced trajectory following
controllers. In the next stages of this work, different
trajectory controllers will be tested with the proposed
algorithm.