A. Bugra Koku

Localization and control of an autonomous orchard vehicle

A new row localization system which uses a laser scanner is proposed.The proposed methodology offers a successful turning between rows of trees.The proposed methodology can run without using a GPS system.The algorithms can be adapted into the real autonomous orchard applications. In this paper we propose a novel model-based control method for an autonomous agricultural vehicle that operates in tree fruit orchards. The method improves path following performance by taking into account the vehicles motion model, including the effects of wheel sideslip, to calculate speed and steering commands. It also generates turn paths that improve visibility of the orchard rows, thus increasing the probability of a successful turn from one row into another, while respecting maximum steering rate limits. The method does not depend on GPS signals for either state estimation or path following, relying instead only on data from a planar laser scanner and wheel and steering encoders. This makes it suitable for real agricultural applications where acquisition cost is key to a farmers decision to invest in new technologies. We show the controllers stability using Lyapunov functions and demonstrate its feasibility in experiments conducted in an orchard-like nursery.

Global urban localization of outdoor mobile robots using satellite images

Localization is one of the major research fields in mobile robotics. With the utilization of satellite images and Monte Carlo localization technique, the global localization of an outdoor mobile robot is studied in this paper. The proposed method employs satellite images downloaded from the Internet to localize the robot iteratively. To accomplish this, the proposed method matches the local laser scanner data with the segmented satellite images. Initial test results conducted on the METU campus are found to be quite promising. Further improvement of this approach has the potential of cutting down not only the operational costs but also the preparation period of the mobile robot enabling researchers to operate their robots in diverse outdoor settings.

Effects of asynchronism and neighborhood size on clustering in self-propelled particle systems

In this study, we analyze the effects of asynchronism and neighborhood size on the collective motion of multi-agent systems. Many studies performed on the collective motion of self propelled particle systems or basically a class of multi-agent systems are modeled to be synchronous. However, in nature and in robotic applications the autonomous agents mostly act asynchronously. Therefore, a model based on the asynchronous actions of agents is developed. The agents/particles are assumed to move with constant speed and asynchronously update their direction of motion based on a nearest-neighbors rule. Based on these rules simulations are performed and the effects of asynchronism and neighborhood size on the clustering performance are investigated.

A new hardware-in-the-loop simulator for control engineering education

As a part of a “lab-at-home” education paradigm for control engineering courses, this paper proposes (and elaborates) a novel hardware-in-the-loop simulator with 3D animation capabilities. The developed software, which can be tailored to simulate any dynamic systems in non-real-time, is designed to work in conjunction with a control hardware. In the paper, the specific application of the software to a graduate-level course is presented within the framework of a final term project involving the control of a satellite tracking antenna. The success of the software (along with the methodology) is rigorously evaluated through the information collected in three academic semesters including the course instructors feedback and the questionnaires filled out by the students.

Global Urban Localization of an Outdoor Mobile Robot with Genetic Algorithms

The localization of mobile robots has been studied rigorously in the past. However, only a few studies have focused on developing specific Genetic Algorithms (GAs) to address the localization problem effectively. In this study; the global urban localization of an outdoor mobile platform is considered with the utilization of the odometer, the laser-rangeq finder measurements and the digital maps created from the relevant satellite images on the Internet. The localization issue is formulated as a constrained optimization problem. The study proposes a GA-based technique to solve the problem at hand efficiently.

A novel command generation method with variable feedrate utilizing FGPA for motor drives

This paper focuses on a novel command generator for servo-motor drives to be used as an integral part of their motion controllers. The method, which incorporates a new data compression algorithm, is capable of generating trajectory data at variable rates. In this paradigm, higher-order differences of a given trajectory (i.e. position) are first computed and thus the resulting data are compressed via the proposed technique. The generation of the commands is carried out according to the feedrate (i.e. the speed along the trajectory) set by the external logic dynamically. The paper discusses the implementation of the method on a Field Programmable Gate Array (FPGA). During implementation Very High Speed Integrated Circuit Hardware Description Language (VHDL) is used rather than using embedded processors on the FPGA chip. The performance of the method is assessed according to the resources used in the FPGA chip on the development board and these results are also compared with the same approach without an interpolator.

A novel soft-computing technique to segment satellite images for mobile robot localization and navigation

Localization of mobile robots has been studied rigorously in the last decade. A number of successful approaches such as Extended Kalman Filter, Markov Localization, and Monte Carlo Localization assume that the map of the environment is originally presented to the robot. However, an important information package like the map of the environment could not be taken for granted in most real- world problems. In this study, a novel technique composed of a combination of Fuzzy C-Means and Fuzzy Neural Network methods is proposed to segment and convert a satellite image into a digital map for outdoor mobile robot applications.

Dynamic modeling and parameter estimation for traction, rolling, and lateral wheel forces to enhance mobile robot trajectory tracking

Studying wheel and ground interaction during motion has the potential to increase the performance of localization, navigation, and trajectory tracking control of a mobile robot. In this paper, a differential mobile robot is modeled in a way that (traction, rolling, and lateral) wheel forces are included in the overall system dynamics. Lateral wheel forces are included in the mathematical model together with traction and rolling forces. A least square parameter estimation process is proposed to estimate the parameters of the wheel forces. In order to implement the proposed methodologies, an experimental setup is used. The setup contains a differentially driven mobile robot, a specially constructed test surface, and a camera system attached at the top of surface for obtaining ground truth. Models having one or more wheel forces are simulated to find the most realistic model. Simulation results are verified by experiments.

Self-learning road detection with stereo vision

It is a hard to solve problem to detect traversable or road regions especially in unstructured roads or paths. In mobile robot applications, robots usually enter these kinds of roads and regions. To successfully complete its mission, it is important to find roads in these environments reliably. In this paper a novel unstructured road detection algorithm with the capability of learning road regions continuously is proposed.

Long-term prediction of hydraulic system dynamics via structured recurrent neural networks

This work presents a methodology for designing neural networks to predict the behavior of nonlinear dynamical systems with the guidance of a priori knowledge on the physical systems. The traditional neural network development techniques are known to have considerable disadvantages including tedious design process, long training periods, and most notably convergence/stability problems for most real world applications. The presented approach, which circumvents such bottlenecks, is especially useful in developing efficient neural network models when full-scale models are not available. This study illustrates the application of the method on a highly nonlinear hydraulic servo-system so to estimate accurately the chamber pressures of its hydraulic piston in extended time periods.