A Robotic Soil Excavator for Truck Loading

Abstract

Hydraulic soil excavators are commonly used in construction sectors and soil removal which are distinguished by high power capabilities and good performance. In this graduation project and for the purpose of automating soil excavation, a robotic soil excavator prototype is developed. It consists of four degree of freedom (4 DoF) with four rigid links connected by four revolute joints. The soil removal requires an expert operator to perform such tasks and consumes time and human power. Consequently, the prototype is conceptualized to achieve a semi-autonomous motion control to dig soil from a given excavation point, carry soil, and finally throw it to a truck loading within a given time duration. Furthermore, it can be operated by an operator remotely. Developing this robotic excavator system involves the machining of mechanical frames, selection of proper electromechanical components, integration of all the parts together. Kinematics and dynamics models are derived for the obtained design to analyze and plan motions, find necessary driving torques and accompanying reaction forces, and to serve as the core of several model-based motion control algorithms. These algorithms are developed and tested using MATLAB and Simulink software packages. The robotic excavator prototype is built following procedures of material selection, design of mechanical structure, and analysis using SOLIDWORKS program. Following that, mechanical components are assembled and tested to satisfy desired functions and given specifications. Each robot joint is equipped with a position sensor and a DC-motor that is controlled via a driver operated in torque mode. The developed robotic soil excavator is considered as an embedded system as it has its own information processing and control unit on board. For this, Raspberry Pi microcomputer is employed to implement a centralized control algorithm that is developed and coded in Python. Experiments show that a trajectory can be generated either through direct or inverse kinematics and can be tracked within acceptable accuracy. Specifically, PD control with gravity compensation is tested in depth. ROS ideas and concepts are examined and implemented. It turned out as discussed in this report that, given the microcontroller chosen with its imposed limitations, relying fully on ROS is not straightforward and requires developing efforts and time beyond the scope of this project.