Preventing the capsize of industrial vehicles: experimental tests on a scaled AGV Benedetto Allotta 2 , Fabio Bartolini 1 , Riccardo Costanzi 1 , Roberto Giusti 1 , Niccolò Monni 1 , Marco Natalini 1 , Luca Pugi 2 , Alessandro Ridolfi 1 1 University of Florence: Dept. of Industrial Engineering via di Santa Marta 3 Florence, Italy roberto.giusti@unifi.it 2 MDM Team S.r.l. via Panconi 12 Pistoia, Italy benedetto.allotta@unifi.it Abstract — The stability of industrial vehicles, such as forklifts and lifters, is very important from a safety point of view: these vehicles are subjected to variable loading conditions and their design is often optimized to privilege handling in narrow spaces instead of stability. In order to study in deep the problem of vehicle capsize prevention and to have the possibility to perform the necessary related experiments, the authors developed a scaled AGV (Automated Guided Vehicle). It is, in particular, a three wheeled differential drive mobile robot (built with low cost commercial components and fast prototyping techniques). The position of the wheels and the loads can be easily modified in order to simulate different vehicle configurations and operating scenarios . The vehicle is controlled through a Texas Instrument C2000 controller, programmable by using Matlab-Simulink Embedded Coder™. In addition, the forces exchanged between the wheels and the ground are monitored using low cost load cells, with miniaturized amplification stages. MEMS three-axial accelerometers and gyros are installed in order to detect inertial loads and to estimate the vehicle pose and, through a proper filtering, the ground slope. The speed and the trajectory of the vehicle can be controlled through the front motorized wheels, driven by speed-controlled drivers. The implemented strategy is able to identify the loading conditions of the vehicle by means of a dedicated algorithm: this algorithm is able to evaluate the position of the center of mass from static measurements that are then further refined when the vehicle is in motion with an adaptive filtering based on the fusion of both static and dynamic measurements. Once the vehicle is in motion, the controller, to prevent the vehicle capsize, is able to limit its forward speed without changing the geometry of the assigned trajectory. In this work, the results of the first testing activities are shown, in order to demonstrate the validity and the effectiveness of the proposed approach. I. INTRODUCTION According to the statistical data available in literature [1], [2], industrial forklifts are often subjected to accidents since the weight and the position of the carried load can vary in a wide range, while human operators can have only a limited knowledge of the load inertial properties and stability conditions. These kind of vehicles are specifically designed to operate in narrow spaces, privileging manoeuvrability more than stability. Furthermore, they are used in warehouses, factories or other industrial buildings where the presence of vehicles, human operators and physical obstacles may increase the accidents probability. Many forklift manufacturers [3] worked and still work to increase the vehicle safety and performances, to meet customer requirements. In particular, from the statistics about mortal accidents, three main causes or typology of events can be identified: 1. vehicle instability: wrong load positioning or inappropriate driver manoeuvres may lead to vehicle instability, causing, e.g., the vehicle to capsize; 2. falling of the load: load stability on forks depends on its correct placement/fixing, related to inertial forces arising during the vehicle motion; 3. vehicle impacting on operators or obstacles. Increasing the level of intelligence and automation can reduce fatality occurrence. Many researchers have developed studies concerning suitable control systems, able to assure the stability of material handling vehicles, [4],[5],[6].