Design and Verification of Control and Tracking Algorithms for Mobile Mini Robot Systems

The topic "autonomous mobile cooperating robot systems" is one of the most seminal technologies. This is indicated by the fact that leading nations like the USA and Japan have adopted it to one of their national research topics. In 2002 the Institute for Machine and Process Automation, the Institute of Computer Technology and the Knowledge-Based Systems Group of the Institute of Information Systems decided to cooperate in research of autonomous mobile robots. The research topic for this proposal can be split up to three subtopics: Position tracking and prediction Path planning Slip control and friction estimation The goal of this proposal is to solve the fundamental control problem, which comes up within the scope of a scientific robot project. The result is used for controlling a mobile mini robot. This mobile mini robot, which was developed in the last years, will be upgraded to a complete autonomous robot. On the topic of position tracking and prediction we will focus on the topic of sensor fusion with acceleration sensors, gyro sensors, yaw rate sensors, encoders, on-board cameras and a global position system as far as it is relevant for controlling a mini robot system. To control a robot system the position of the robot relative to its environment is required. In the used testing bed the movement of the robot, which has to be controlled, as well as the objects moving in a near surrounding to the robot, are detected by the help of on-board sensors. Based on multiple sensor information the actual and future positions of the moving objects and the environment are generated. The control algorithm itself is based on a layer model, whereas this project focuses at the lowest control levels. These are the controlling of the robot to a target position by generating a trajectory and control the robot along this trajectory. During this work algorithms for generating such a trajectory will be developed whereas the path is subject to boundary conditions. These boundary conditions can be an area wherein the robot should stay or obstacles the robot has to avoid. The result of the path planning is a trajectory for the robot. To control the robot along a trajectory slip control mechanisms will be developed by using acceleration sensors in addition to the encoders, which are mounted on the DC motors.

The fundamental task in autonomous mobile robot motion control is to make the robot move to a specified goal position. This task is divided into a number of sub-tasks, which are executed in a hierarchical manner. Firstly, it is necessary to plan the motion to the goal, i.e. a reference trajectory or path is generated (`Trajectory generation`). Secondly, the robot needs to keep track of its position while in motion (`Navigation`). Thirdly, the robot needs to be driven such that the reference trajectory is accurately executed (`Tracking control`). In the course of this work novel concepts for all three sub-tasks were developed and tested: For trajectory generation, various criteria are weighed against each other. Side-accelerations during sharp turns, time to reach the goal, total arc length and smoothness of the curve are simultaneously optimised, while hard constraints such as maximum speed, minimum turning radius (i.e. maximum steering angle), maximum acceleration and minimum safety distance to obstacles are accurately respected. The unified mathematical formulation of this problem and its numerical solution constitutes a novel achievement. The developed navigation algorithm is derived from an existing idea but is extended to a more comprehensive concept. The idea is to navigate based on the wheel revolutions, while constantly monitoring the wheel slip. In case of excessive slip, which makes navigation based on wheel revolutions unreliable, inertial sensors such as acceleration and gyro sensors are substituted. Monitoring also enables control measures to limit the slip. The application of predictive control to trajectory tracking control constitutes an innovative way to deal with certain mathematical problems in mobile robot control. The robot is capable of optimising its control inputs based on a prediction of its movement. This resembles closely the behaviour of a human driver. Possible applications of the presented algorithms are autonomous industrial transport devices, autonomous vehicles for deployment to dangerous environments in case of natural disasters or driver-assist systems in automotive applications.