Conversion of a standard vehicle into a Unmanned Ground Vehicle. Focus on Sensor fusion and Extended Kalman filter

Nowadays, there are several fields where the need to delegate human actions to robotics is more and more increasing. One of the major development deals with Unmanned Ground Vehicles (UGVs), which are defined as ground vehicles without a human presence on board. Their main purpose is to physically replace the human driver in situations where his presence can be disadvantageous or even dangerous. Currently, the main applications are related to military and rescue environments, or space exploration. This project was realized in cooperation with the AMET Srl company and consists in the conversion of a standard vehicle into a UGV, equipped with a remote-control station and assisted by a local control. The idea is that the remote driver, through a data entry, provides the needed inputs to the vehicle. The control is responsible for processing the inputs and converting them into commands to be acted to the vehicle. Therefore, output variables are collected by sensors and sent back to the remote station. Due to teleoperation delays in the communication channel, remote drivability might not be always guaranteed. In this sense, the local control is needed to reject local disturbances, compensate for the communication delay by taking immediate decisions while waiting for the driver’s response. In order to design the control, actuators and sensors, a Model Based Design (MBD) approach has been used, where a Plant-Control-Host (PKh) architecture is needed to clearly distinguish the different parts of a mechatronic system and thus to support all the phases of a V-cycle. This work is focused on the Model-in-the-loop (MIL), where both Plant and Control are simulated in a modelling framework. Future developments could be related to Rapid Control Prototyping (RCP), Software-In-the-loop (SIL), and then to Hardware-In-the-loop (HIL) tests. Specifically, the Plant includes the vehicle and the scenario models, and it is entirely modelled in TruckSim, a software able to analyse vehicle dynamics, predicting vehicle behaviour on the bases of driver controls; the Control and the Host are modelled in a MATLAB & Simulink framework. TruckSim is produced by an American company, Mechanical Simulation Corporation, and is distributed in Italy by AMET. This thesis has been focused on Sensor fusion, with the implementation of an Extended Kalman filter (EKF), aimed at providing an accurate estimation of variables that are not directly measurable or affected by uncertainty. The EKF is needed for teleoperated guide purposes, making controls more robust; as a matter of fact, filter’s outputs are exploited as input of Advanced Driver Assistance Systems (ADAS). In the end, an Adaptive Cruise Control (ACC) has been implemented to test the EKF performance.