Affiliation:
1. Robotics Innovation Center , DFKI GmbH (German Research Center for Artificial Intelligence) , Bremen , Germany
2. Faculty of Mathematics and Computer Science , University of Bremen, Robotics Group , Bremen , Germany
Abstract
Abstract
The Autonomous Rough Terrain Excavator Robot (ARTER) is a retrofitted walking excavator developed for remote and autonomous operations in environments hostile to humans. This work highlights the key developments related to this robot: system design, terrain adaption controller, and high-level process controller. The original walking excavator is retrofitted with sensors, hydraulic valves, computation devices, etc., to automate it. The terrain adaption controller, which adapts the wheels automatically to the underlying uneven terrain, is implemented using deep reinforcement learning. The tasks for the robot are complex and require switching between autonomy and remote operations. Hence, a custom high-level process controller, based on behavior trees, which helps the operator control complex tasks for the robot, is developed. The remote control and autonomous behaviors of the robot are evaluated for realistic scenarios performed in a test environment.
Subject
Electrical and Electronic Engineering,Computer Science Applications,Control and Systems Engineering