Elsevier

Applied Soft Computing

Volume 11, Issue 1, January 2011, Pages 1303-1309
Applied Soft Computing

Automatic lateral control for unmanned vehicles via genetic algorithms

https://doi.org/10.1016/j.asoc.2010.04.003Get rights and content

Abstract

It is known that the techniques under the topic of Soft Computing have a strong capability of learning and cognition, as well as a good tolerance to uncertainty and imprecision. Due to these properties they can be applied successfully to Intelligent Vehicle Systems; ITS is a broad range of technologies and techniques that hold answers to many transportation problems. The unmanned control of the steering wheel of a vehicle is one of the most important challenges facing researchers in this area. This paper presents a method to adjust automatically a fuzzy controller to manage the steering wheel of a mass-produced vehicle; to reach it, information about the car state while a human driver is handling the car is taken and used to adjust, via iterative genetic algorithms an appropriated fuzzy controller. To evaluate the obtained controllers, it will be considered the performance obtained in the track following task, as well as the smoothness of the driving carried out.

Introduction

Intelligent Transportation Systems (ITS) apply information and communication techniques in order to achieve safe and efficient driving. In the automotive industry, ITS mainly is used to provide information to the driver and, in some cases, they are connected to the car actuators, attempting to minimize injuries and to prevent collisions [1].

The work described in this paper has been carried out at the AUTOPIA Program of the Centre of Automatic and Robotic, a part of the Spanish Council for Scientific Research (CSIC), which is focused on the area of autonomous vehicles. More concretely, the aim of the present work is the automatic control of the steering wheel (also known as lateral control) of a mass-produced vehicle, although another area of the research carried out by the AUTOPIA Program is longitudinal control, the control of the vehicle’s speed and its adaptation to road features, using the throttle and the brake pedal as needed [2], [3]. Research projects similar to AUTOPIA exist in different countries [4], [5], [6], [7].

Vehicles used by the AUTOPIA program are equipped with instrumentation and software necessary to perform their autonomous management, since an autonomous car must control some or all of its functions without external intervention [8], [9], [10]. To carry out the present work, they have been required: An asphalt set of test roads called ZOCO (acronym for Driving Zone), differential GPS coverage supplied by a high-precision GPS base station and a wireless LAN for differential correction generation and transmission and a mass-produced Citron C3 Pluriel equipped with a DGPS, actuators for the steering wheel and pedals, and a computer connected to the communications network. There is long way until an implementation of full automatic-steering control comes on the market [11]. Due to that, researchers around the entire world focus their efforts in the study, implementation and validation of this kind of systems as is showed in the famous driverless cars competition sponsored by the Defense Advanced Research Projects Agency1 (DARPA), the central research organization of the United States Department of Defense. DARPA has sponsored three competitions in the area of autonomous vehicles.

Classical control techniques are the usual way to manage complex systems such as the steering wheel of a car [12], [13]. Other way is the use of artificial intelligence techniques; these methods are specially indicated when we try to emulate human control actions, such as human car driving. Fuzzy logic [14] has become a particularly widely used methodological approach to these tasks since Sugeno’s work on fuzzy control in 1985 and 1987 [15], [16]. Fuzzy systems arose from the desire to describe complex systems linguistically, and fuzzy controllers allow a human approach to control design without the demand for knowledge of mathematical modeling of more conventional control design methods [17]. Usually, generation of the membership functions and rule base that define a fuzzy controller is a task mainly done either iteratively, by trial-and-error, or by expert knowledge. A task like this one is a natural candidate to be solved by means automatic techniques from the artificial intelligence field.

Genetic algorithms (GA) are general purpose search algorithms which use principles inspired by natural genetic populations to evolve solutions to problems [18], [19]. The basic idea is to maintain a population of chromosomes, which represent candidate solutions that evolves over time through a process of competition and controlled variation. Each chromosome in the population has an associated fitness to determine which chromosomes are used to form new ones in the competition process, which is called selection. The new ones are created using genetic operators such as crossover and mutation. GA have been used in literature in task of learn or optimize fuzzy controllers [20], [21], [22].

The ORBEX [23] (Spanish acronym for Fuzzy Experimental Computer) fuzzy development environment has been used for the definition and implementation of fuzzy controllers. With ORBEX, several ways of driving can be defined to emulate different types of drivers: calm, quick, brusque, etc., or to adapt the driving to traffic conditions: platoons, overtaking, etc. These strategies can be defined and implemented by means of ifthen rules in almost natural language [24].

This paper approaches the topic of using of genetic algorithms in the design of fuzzy logic controllers a real world application: a mass-produced vehicle steering wheel management. After a first part of capture of information about a human driver handling, that information is processed in order to get relevant information about the attitude of the driver. Once done this, a system able to get the information and return an appropriate fuzzy controller has been created, via application of an iterative genetic algorithm. The GA has been implemented in two iterative phases, the first one will improve the membership functions and the second one the rule base of the fuzzy controller. Finally, obtained controllers have been tested in a private and experimental area to verify that they show a similar behavior to the shown by human driver.

Section snippets

Information capture and processing

Generated fuzzy controllers will use two inputs obtained from the match between the GPS positioning information and the reference route, defined as GPS digital cartography. The input variables are: the lateral error and angular error. The lateral error represents the distance of the current car position to the theoretical car position if it was on the desired trajectory, the reference line; its values can take any value (, +). The angular error is the angle shaped by the reference line and

Fuzzy controller optimization

It is natural to think that the fuzzy controller for managing the steering wheel of a car must be symmetric; this would mean that the action taken when the inputs have a certain value must be inverse to the action taken when the inputs have inverse values. However, it is not the case because the aim of this work is to reproduce human driving so the controller does not have to be symmetric. Therefore, the system must be able to distinguish, for example, a driver who drives too far to the right

Experimentation and results

In the presented method we can observe the following aspects that have to be considered when parameters for the experiments are configured: there is always a copy of the best chromosome found in the population; before the improvement of each part of the controller (MF and RB) a new population is created and the improvement of each part of the fuzzy controller is done separately. Bearing in mind the characteristics of the method, the parameters must fulfill the following premises:

  • NMF/RB must be

Conclusions and future works

This work has presented an automatic adjustment of fuzzy controllers designed to automatically manage the steering wheel of a mass-produced vehicle which was equipped with the necessary instrumentation and software. To achieve this, an iterative genetic algorithm based method was implemented, capable of iteratively adjusting the membership functions and rule base which define a fuzzy controller. The method applied genetic algorithms with some constraints applied to the controllers to guarantee

References (26)

  • L.A. Zadeh

    Fuzzy sets

    Information and Control

    (1965)
  • F. Herrera et al.

    Tuning fuzzy logic controllers by genetic algorithms

    International Journal of Approximate Reasoning

    (1995)
  • M. Mucientes et al.

    Design of a fuzzy controller in mobile robotics using genetic algorithms

    Applied Soft Computing

    (2007)
  • W.D. Jones

    Keeping cars from crashing

    IEEE Spectrum

    (2001)
  • J.E. Naranjo, C. González, J. Reviejo, R. García, T. de Pedro, Adaptive fuzzy control for inter-vehicle gap keeping,...
  • V. Milanés et al.

    Autonomous vehicle based in cooperative GPS and inertial systems

    Robotica

    (2008)
  • M. Bursa

    Big names invest heavily in advanced ITS technology

    ISATA Magazine

    (2000)
  • J.M. Blosseville et al.

    The French program: La route automatisée

    IEEE Intelligent Systems

    (2000)
  • Literature Review of ADAS/AVG (AVCSS) for Japan and United States,...
  • R. Bishop

    Intelligent vehicle applications worldwide

    IEEE Intelligent Systems

    (2000)
  • J.E. Naranjo et al.

    Using fuzzy logic in automated vehicle control

    IEEE Intelligent Systems

    (2007)
  • R. García et al.

    Automatic car drivers

  • M.A. Sotelo et al.

    Vehicle fuzzy driving based on DGPS and vision

  • Cited by (0)

    This work was supported by the Plan Nacional under the project Tránsito (TRA2008-06602-C03-01), by the Comisión Interministerial de Ciencia y tecnología under the project GUIADE (Ministerio de Fomento T9/08), by the Consorcio Estratégico Nacional en Investigación Técnica under the project Marta (CENIT-20072006) and the Ministerio de Ciencia e Innovación under the project CityElec (PS-370000-2009-4).

    View full text