I. Introduction
Legged robotics has potential advantages in terms of mobility and versatility as compared to tracked or wheeled vehicles. So far, the technological complexity to build and control such vehicles has prevented these systems from being applied in real world scenarios and only few teams managed to develop machines that work beyond laboratory test-bench settings. With major advances over the recent years, pushed by various large scale research programs or investment from industry, our community is about to overcome the last technical hurdles and make legged robots available for real world applications. Most prominently, the DARPA Robotics Challenge (DRC) brought together some of the best research groups in the field of humanoid robots to successfully use such machines in a disaster mitigation scenario [1]. Since the scenario is very close to reality, all teams were forced to massively invest in hardware development to improve not only versatility but also reliability and ruggedness of the robots. These developments resulted in many high-performance machines like ATLAS [2], Valkyrie [3], DRC Hubo [4], HRP2+ [5], Walkman and others, most of them based on earlier robot versions. This new generation of humanoid robots commonly feature some sort of force or torque control - either by integrated load cells in the joints or at the end-effector, or by a series elasticity in every actuator. This allows them to properly control interaction forces with the environment and hence to balance the system or manipulate the environment.
ANYmal, an autonomous quadrupedal robot for rough terrain operation