Identification Methods and Simulation Modeling of a small UGV for Indoor Applications

The last few years have seen a growing interest in the development of intelligent machines capable of moving autonomously in space and being aware of their surroundings. Their great potential makes them ideal for the most varied fields of application: agriculture, manufacturing, land and aerial surveillance, naval operations, commercial transport, space exploration. A focus on mobile robots can be observed, since they are systems able to integrate technologies related to sensing, information processing, movement on wheels, obstacle avoidance technique and Artificial Intelligence. Their versatility makes them the test bench suitable not only for testing high level solutions of Guidance, Control and Navigation techniques (GNC), but also at a strictly mechanical level with the possibility of testing configurations that include robotic arms, payloads for environmental analysis, hybrid motion mechanisms. In a space framework, rovers can be considered as evolution of these mobile robots: these robotic systems for the exploration of planetary surfaces are designed to move independently in unknown environments, to conduct analysis on the ground that can identify composition and properties, to photograph the surrounding environment and allow its study on Earth. Given the cost associated with space operations and the rare possibility of repair, the reliability of such systems must be of the highest level. It is therefore clear the importance of testing their effectiveness in similar situations to those in which they will operate on the designed mission. The tests are possible thanks to numerical simulations. This thesis is addressed to the system identification of a mobile robot to obtain a model that can be used in a simulation environment. The advantage offered by simulation is linked to the possibility of carrying out tests that do not damage the real device and do not require its continuous use. Devastator robotic platform, a tracked mobile robot controlled by an on-board companion computer, is the reference used for identification and experimental testing. The analysis of the DC motors behavior is first performed with both a lumped parameter approach and a data-driven approach. A comparison between the two methodologies is carried out. A kinematic model is then adopted for the reproduction of the robot motion, but the identification of the inertial parameters of the system has also been performed as a basis for future development and integration of a dynamic model. The thus obtained model is recreated in MATLAB/Simulink environment. A controller and a path planning algorithm are also implemented to test and validate the behavior of the plant. The selected control strategy is based on a Proportional–Integral–Derivative or PID controller while a simple but effective Artificial Potential Field strategy is adopted for path planning. Their performances are subsequently established through SimulinkROS/Gazebo co-simulation; Gazebo is in fact a simulation tool for algorithms, configurations and control strategies testing in a realistic scenario. Finally, they are translated into executable code supported by on-board systems thanks to the MATLAB Code Generation functions and tested in indoor applications.