Grupo de Inteligencia Artificial

1
Grupo de Inteligencia Artificial
UNIVERSIDAD TECNOLÓGICA NACIONAL
Facultad Regional Buenos Aires
Argentina
2
Grupo de Inteligencia Artificial
Grupo de Investigación Origen 1986 Investigación
y desarrollo de temas relacionados con la IA y
nuevas tecnologías Año 2000 necesidad de
aplicación de los temas desarrollados
3
Plataforma Robot
4
Plataforma Robot
5
Red interna
6
Módulos de la plataforma Robot

• Módulo de Reconocimiento de Imagen (interactua
  con el Radar de Ultrasonido)
• Módulo de Reconocimiento de Voz (para reconocer
  comandos básicos)
• Módulo de Lógica Difusa (para la corrección de
  posición por odometría y detección de Hitos)

7
UNIVERSIDAD TECNOLÓGICA NACIONAL
Facultad Regional Buenos Aires
GRUPO DE INTELIGENCIA ARTIFICIAL (GIA)
Route planning for vehicle autonomous
navigation, based on geometrical regions. Part
II Multiple approach points
8
UNIVERSIDAD TECNOLÓGICA NACIONAL
Facultad Regional Buenos Aires
GRUPO DE INTELIGENCIA ARTIFICIAL (GIA)
L. M. Di Matteo, A. C. Mangone, M. L.
Muzzio, C. Verrastro
9
Introduction
Algorithmic methods based on potential fields
have demonstrated to be the most effective ones
(Kathib at al., 1998) (Borenstein y Koren,
1991b). These techniques are very stable against
sensing errors and easy to implement, but these
have some difficulties when obstacles are very
near and the obstacle free paths are very thin.
10
Introduction
There are several control techniques based on
artificial intelligence (Patino and Carelli,
2003) that have to lead with structured or
dynamic environments.
Others works model the environment on a grid
based map (Pereiro and Verrastro, 2003) (Weigl at
al, 1993) where each cell is occupied by an
obstacle or not. This model is very useful to
process search algorithms in optimum path
searching.
11
Introduction

• Some vehicles have big inertial factors and large
  turn ratio.
• Others move very fast and have poor braking
  systems
• It can be applied to airplanes, ships, etc.

12
Objectives
The main task of this algorithm is to determine a
trajectory to achieve a defined point, avoiding
obstacles and arriving to that point with a
specified entry angle. Current and target
position must be given, as well as the
environment knowledge.
13
Objectives
The feature to obtain is to arrive with a high
precision to the goal point, that is to say, to
reach the goal point with a very low error in the
entry angle. This algorithm is complementary to
the route planner.
14
Environment Modelling
15
General Description
16
General Description
Multiples approach points are generated So, it
is needed several approximation
radius Successive approximation radius are
generated, each radius grows up with an
exponential law.
17
General Description
Each approach point is replaced with a
bidimensional band of error, defined as the
distance to the corresponding approach
point. The error distance changes depending on
the approach point considered. For outer
approximation radius the error distance increases.
18
General Description
19
General Description
Different geometrical regions are defined. These
regions are composed with the junction area
between approximation radius, plus the error band
of the outer radius, minus the error band of the
inner radius.
20
General Description
21
General Description
The algorithm behaves as follows It returns the
following coordinate to reach. The mobile must go
to the approach point located in the next region,
that is, the region nearer to the goal point.
22
General Description
When mobile arrives to the region, the algorithm
changes the current target coordinate by the
coordinate of the approach point located in the
next region, an so on. This loop is repeated till
goal point is reached.
23
General Description
24
General Description
This type of trajectory is expected.
25
Experimental Results
26
Conclusions
As seen in the experimental results, the
algorithm proposed achieves its objective
successfully, the simulation demonstrated that it
fits to vehicles with inertial troubles. The
mobile reaches the goal point with a low error in
the entry angle.
27
References
Borenstein J. and Y. Koren, The Vector Field
Histogram Fast Obstacle Avoidance For Mobile
Robots, IEEE Journal of Robotics and Automation,
7, No 3, 278-288 (1991a). Borenstein J. and Y.
Koren, Potential Field Methods and Their
Inherent Limitations for Mobile Robot
Navigation, Proceedings of the IEEE-ICRA,
Sacramento, California, 1398-1404 (1991b).
Khatib, O. Real-time obstacle avoidance for
manipulators and mobile robots Proceedings
IEEE-ICRA, St. Louis MO, 500-505 (1985). Patiño
D. and Carelli R. Adaptive Critic Design-Based
Optimal Control For Mobile Robots Navigation, X
RPIC Proceedings, San Nicolás, Bs. As., 503-507
(2003). Pereiro F. and Verrastro C. Sistema de
Comando y Navegación para Robot Móvil con
Arquitectura Distriuida X RPIC Proceedings, San
Nicolás, Bs. As., 565-569 (2003). Weigl M.,
Siemiatkowska B., Siroski K., Borowski A.
Grid-Based mapping for autonomous mobile robot,
Robotic and Autonomous Systems, Amsterdam,
Holland, (1993). Orqueda O. and Agamennoni O.
Motion Planning and control of Autonomous robots
I Generalized Potential Field Functions X
RPIC Proceedings, San Nicolás, Bs. As., 541-546
(2003). Hwang, Y.K. and N. Ahuja. Gross Motion
Planning A Survey, ACM Computing Surveys,
24(3), 219-291, (1992). Latombe, J. C. Robot
Motion Planning, Kluwer Academic Pub. Boston
(1991) Alberino S., Folino P. and Verrastro C.
Variante en el algoritmo PID para evitar el uso
de un generador de trayectoria trapezoidal X
RPIC Proceedings, San Nicolás, Bs. As., 659-663
(2003).
28
Grupo de Inteligencia Artificial
www.secyt.frba.utn.edu.ar/gia/