Robots of Roboy Milestones towards a human body.
The next Roboy
The Roboy team is currently exploring the possible advancements that could be implemented on the next Roboy: Roboy 2.0.
Mobility, communication and Virtual Reality (VR) are the ongoing fields of improvement that the team is working on. The “Stand-up” team is building the required technology to make Roboy stand on his own legs. The “Hi there” team is further developing Roboy so that he can interact naturally with people. The “Virtual Roboy” team is currently working on using today’s VR technology to enable anyone to feel like Roboy feels.
Here is more info about the current teams’ work.
Various projects have stemmed from the research of Roboy 2.0:
The MyoArm stems from the EU funded MyoRobotics project. It is based on the same artificial muscle systems developed for Roboy. It is modelled after the human shoulder and arm complex: 11 compliant or ‘giving’ muscles move the shoulder and the attached arm. Just like in our body, each muscle is in a configuration opposing at least one other motor. The compliance in the system allows it (just like us humans) to store energy and release it very fast – perfect for throwing a ball. However, the complex interaction between muscle units in the shoulder, where 9 muscles control three degrees of freedom (up and down, left and right, rotation) poses a variety of interesting control challenges. Just like our brain uses its neural network to control complexities like this, the research team at the Technical University of Munich develops artificial neural networks to control the MyoArm.
A unique combination, taking the best of Roboy and the best of Myorobotics, building new legs with the goal to finally have them stand. Developed in collaboration with University of Zurich and QUO AG, they were the first step of integrating the modular Myorobotics with Roboy’s technology. For this a completely new hip, ankle and feet have been developed. The hip is compatible with the original Roboy. But that’s a story for the future.
PaBiRoboy (Partial-anthropomimetic Bipedal Roboy) is inspired by NaBiRoS developed by RoMeLa at UCLA. The aim is to build a robot that can only walk in one plane to make the control problem less complex. In comparison to NaBiRos, which uses directly actuated joints, PaBiRoboy is constructed on anthropomimetic principles, using muscles to move his joints. As with the human body, each joint has two muscles acting on it, one antagonist and one protagonist. PaBiRoboy is being developed as a stepping stone for enabling the Roboy legs to walk.
HTC vive lighthouse custom tracking
Our goal is to make Roboy balance and walk. For this we need accurate tracking. We decided to replicate HTC lighthouse tracking for our purposes and in November 2016 a fascinating journey began.
For the first prototypes we were using openvr/steamvr. Steamvr comes in prebuild libraries and you need special LD_LIBRARY_PATH settings for succesfully running it. This conflicts with our ros kinetic builds and is very messy afterall. Then after one of the recent updates steamvr stopped working on our computers. This was actually the reason we started digging into how HTC might have solved the tracking.At first we tried decoding the signals from the sensors using Intel Edison and MKR1000, which in case of the Edison turned out to be impossible and in case of the MKR was limited to a small amount of sensors. For the Edison the hardware interrupts were not handled fast enough. This is due to a threaded interrupt system. We also tried using the MCU, which wasn’t fit for the job either. The MKR was simply overwhelmed by all the interrupts. We disassembled one of the HTC vive controllers for getting our hands on those sensors. We noticed the HTC controllers were using an ICE fgpa. So we thought if they use it, there must be a reason.Then soldered VCC, GND and Signal copper cables (0.1 mm), using enough flux. And covert the sensor with a bit of glue to protect it from accidental damage. In the previous prototype we were routing all signal cables coming from the sensors in parallel. This turned out to be a bad idea. Because of induction the signals pollute each other. In the vive controller, they deal with it by isolating the signal with VCC and GND. So thats what we are also doing.
More information about this can be found here.
Human have the best evoluted hand comparing with the other species. Building a mechanical hand after human’s hand is a challenging and interesting thing.
Roboy Hand is designed based on the genera idea of Roboy, whose fingers and palm are all driven with tendons and motors from the arm. Because of the complexity of human hand, lots of simplifications (like reducing the bone numbers and integrated mechanical joints) are required. At end a control system for about more than 23 motors should be build and the robotic hand can be widely applied.
The 3D Printing technology helps building a prototype with complicate structure and consideration and modeling of the complex tendon system and bone structure inspire biomechanical and clinical innovations and applications. Ideas like neural network and genetic algorithm can be also get applied with the researching and developing robotic hand.
Collaboration in Research
The Human Brain Project (HBP)
Roboy is part of the Human Brain Project, in the sub-project (SP) 10 of neurorobotics.
One of the goals of the Neurorobotics SP is to translate virtual robots and brain-derived controllers to physical prototypes to contribute to future robotics research. To this end, results from brain research and neurorobotics are transferred to future modular robots and state-of-the-art embedded systems. In particular, the teams investigate closed-loop models using neuromorphic hardware and physical robotics hardware to develop embedded neurorobotics systems. For the physical robots, there will be a strong focus on innovative soft- and muscle-tendon driven robots that are more suitable for human environments, like the „Roboy“ and the „MyoArm“. Roboy acts as an ambassador for safe, easily accessible and open-source, human-inspired robotics. Roboy is not only a scientific challenge, but also a very well known robot and regularly advertises the HBP and open robotics initiatives, such as the MyoRobotics hardware at his many public appearences. Within the HBP Roboy is the first anthropomimetic robot available in the Neurorobotics simulation platform. This enables researchers to develop neuro-inspired algorithms independently within the platform on an existing bio-inspired robot and potentially validate them on the real robot afterwards. The MyoArm is a showcase of the usefulness of neuro-inspired algorithms for controlling highly complex system. Being based on open hardware and the de-facto standard of open-source robotic research software, it offers researchers the possibility to easily build experiments and systems.
MyoCDPR is a research project in cooperation with the Chinese University of Hong Kong. The aim of MyoCDPR is to develop an open-source ROS package for the real-time analysis and control of Tendon or Cable-Driven Parallel Robots (CDPRs) actuated by Myomuscle actuator units, similar to Roboy.
Cable/Tendon-driven parallel robots are a class of parallel robots where tendons replace the rigid links. In recent years, this type of robots have attracted great interest due to their very unique advantages: high payload to weight ratio since the actuating motors are fixed on the base frame; potential to span large distances as tendons can be attached to base frame far away from the end-effector; potential for reconfigurability of the attachment points of the cables to change the robot properties without developing a new manipulator.
These advantages have resulted in many interesting applications that are unique to the use of CDPRs: high acceleration or high payload manipulation, search & rescue, SkyCam, large dimension radio telescopes, construction, airplane maintenance, sandblasting of large structures and bio-inspired robots.
One unique characteristic of CDPRs is that the tendons can only be actuated in tension and not compression, resulting in the need of actuation redundancy. This creates challenges in the kinematics, dynamics, workspace analysis, control, and design of CDPRs. In the existing research on CDPRs, both in simulation and experimentation, two interesting traits can be observed in the research methodology:
1) developed analyses techniques are typically only performed one single tendon robot designed by a particular research group;
2) hardware and software for CDPRs are typically not shared and must be re-developed between different research groups.
This inevitably results in “re-inventing the wheel” and inhibiting the advance of CDPR research.
With this motivation, the goal of this project is to develop a combined open-source hardware (Myomuscle) and software (open-source MATLAB software, Cable-Robot Analysis and Simulation Platform for Researchers – CASPR) modular tendon robot actuators for developing and studying arbitrarily structured CDPRs (MyoCDPR).
The development of Roboy (mechanic and software) is conducted Open Source. This means that all expertise, ideas and inventions do not belong to one specific entity, and everyone will has the chance to advance Roboy’s technology. Roboy is the ideal initial spark to trigger the work towards a generic anthropomimetic research platform.
All of the code and CAD files are freely available on github under a very permissive license (BSD 3.0 and CC-BY 4.0).
The parts required to build Roboy are kept in our sponsored aligni.com instance at roboy.open-aligni.com.
The documentation of the work of the current individual teams is available on our team development space.
As Roboy is under heavy development, if you would like to have access, learn how to build your own, or contribute please contact gro.y1503358112obor@1503358112maet1503358112. Alternatively, simply show up to our weekly team meetings on Wednesdays at 10 AM in the Roboy / Tech Talents room, Lichtenbergstraße 6, 85748 Garching, Germany.