Locokit III: A versatile robotic platform to study embodied locomotion

In the past, researchers have had an interest in understanding the hidden secrets of locomotion. To a large extend, we are able to analyze animal locomotion by using high speed cameras, kinematic tracking systems, and ground reaction force measurement devices. This allows us to see what exactly goes on when an animal or human walks, runs, or climbs; thereby gaining more insights into how nature solves the task. However, with this approach several challenges rises. The first challenge is, that re-engineering the mechanical structure of muscles, bones, ligaments, etc. which nature has evolved to almost perfection is very difficult to redo in the world of engineering. Most often we use electric motors where animals uses muscles which are completely different ways of inducing energy into the robot. In the joints, where nature uses ligaments and cartilage, the most used alternative for engineers are bearings and the transfer of energy through ligaments is also a challenging task to mimic in robots. Alternative ways of creating actuation, such as artificial muscles, pneumatic, are also becoming popular. However, they create other problems such as energy efficiency, size, weight, control, etc. Most often the actuation is connected directly to the degree of freedom the actuator has to control in order to make things as stable and easy to control as possible. Another point, which is just as challenging, is the control side. Though we are able to see and study how an animal or human behaves in a certain situation, the challenge of how to generalize that knowledge into something that can be used to control a robot continues to be a challenge. The problem is that two situations are never the same, so just making a copy is not possible. The controller needs to be able to adapt to both the morphology of the robot and the environment. Thought this is of cause common knowledge, and research have been working towards a general controller for years, the physical toolset for testing these systems is still limited. Thus, the majority of research employs simulation. Although simulation can be a very good and efficient tool, the reality gab remains a problems. On the other hand, working with robots can be quite time consuming, costly and annoying not to mention that it requires other competences than what control engineers and biologists might have. The use of a physical robot would always produce correct results in the sense that the results are based on the robot moving around in a physical environment and not inside a simulator. As an alternative to simulations and other toolkits such as Allbot [3], Lynxmotion [4], Vex Robotics [5] etc. Though these toolkits provide abilities to do legged locomotion, to the best of our knowledge, they do not provide the ability to study dynamic locomotion in both walking and running with multiple gaits as LocoKit II [1] [2].