Roboy Open Topics

Roboy Student Team works on developing the tendon-driven humanoid robody Roboy 3.0 and its next generations. Robody is a remote robotic body, controlled at all times by a human operator using VR gear.

All code, results & scientific findings produced by the Roboy Student Team are open-sourced and, if possible, published. Topics of interest cover biologically-inspired actuation, simulation, sensing, control, telepresence and XR, where a wide range of skills in mechanics, electronics & software engineering can be applied.

High-level reasoning approaches, typical for autonomous robots, are out of scope.

Here we collect all ideas for potential Roboy thesis, lab course, etc.

Field	Priority	Partner	Description	References

Field	Priority	Partner	Description	References
Control	HIGH / ONGOING		Mujoco simulation of a tendon-driven musculo-skeletal robot Roboy Research & apply state of the art methods to narrow hardware-simulation gap with goal to open-source the simulation model for the community and research.
	HIGH / ONGOING	MAE, Chinese University of Hong Kong Valero Lab, University of Southern California	Force control of a tendon-driven musculo-skeletal robotic arm Implement kinematic control for Roboy 3.0 (simulation & hardware). Model-based and model-free methods will be taken into account.	Possible collab:
	HIGH	TUM FAR	Wrist control Implement kinematic control and operator mapping for the new stronger wrist of Roboy 3.0
Sensing	MEDIUM	University of Leeds	Soft tactile sensors for Roboy In the scope of this project the team will be developing tactile sensors (aka skin) for Roboy based on the advancements from the University of Leeds, who will be co-supervising the project. Topics cover sensor modification sensor manufacturing data read-out and interpretation	Partner’s papers: Adjustable Compliance Soft Sensor via an Elastically Inflatable Fluidic Dome https://eprints.whiterose.ac.uk/168371/1/SITS_Optimization.pdf
	HIGH		“Manual” Robody Control Explore the control of the Robody using haptic gloves Map the Hand’s state to the robody’s hand Explore ways to switch between operating the robody and navigating the menu Make use of the glove’s haptic feedback	https://www.bhaptics.com/shop/tactglove
	HIGH completed		Facial expression transfer and mapping between the operator and the avatar during telepresense This project covers: recognition of the operator’s emotions while wearing a VR headset (currently Meta Quest Pro) design of new animations for the existing Roboy Face (Unity project) facial expression operator-avatar mapping	Possible HW addition: https://www.youtube.com/watch?v=4_EIqmIWn_Y
	HIGH / ONGOING		Extra-human sensing for Roboy Avatar: infrared, ultrasound, magnetic field Telepresence opens a wide range of possibilities to extend the human sensing capabilities. For example, adding an infrared camera on the avatar and fusing the its image with the RGB camera stream will already give the operator extra-human sensing which will be invaluable in situations like search and rescue operations.	Detailed description:
	HIGH completed		Contactless vitals monitoring Use the cameras to find out if a person is standing/sitting or lying on the floor Use other sensors (e.g. 60 GHz radar) to establish the persons vital data - e.g. body temperature, breathing rate, heart rate, blood pressure, …
VR/AR	HIGH / ONGOING	Infineon	Immersive vision stream for telepresence using ToF / 360 deg camera Currently the vision data in telepresence is using a stereocamera. The field of view it provides, however, is rather limited. Therefore, we seek to augment the RGB camera stream with a long-range (9m) time-of-flight sensor and/or a 360 degrees camera to improve operator’s situational awareness.	Detailed description:
	LOW	Nvidia	Telepresence through Nvidia’s CloudXR & Omniverse This project aims to explore the potential of using the latest Nvidia’s tools for telepresence - as a data transmission and cloud processing layer. These tools will allow to move computationally expensive tasks to the cloud - e.g. video stream super-resoluiton, detailed rendering and simulation, etc. Expert support by the partner.	Develop on NVIDIA Omniverse Platform NVIDIA CloudXR
	MEDIUM		AR Roboy Developing an AR overlay of the Roboy Avatar.	Potential HW: https://www.nreal.ai/light
	LOW		Neural style transfer during telepresence This artistic project will allow the operator to see the reality in a new way through Roboy eyes inspired by Vincent van Gogh or maybe Simpsons. Here we aim to apply real-time time-coherent neural style transfer on the avatar camera stream.	https://blog.unity.com/technology/real-time-style-transfer-in-unity-using-deep-neural-networks
Mechatronics	completed		Standing Roboy Avatar Currently Roboy 3.0 is sitting on an electric wheelchair. The goal of this project is to make Roboy stand up by rigidly mounting him to a self-balancing platform (e.g. Segway).
	MEDIUM		Next generation muscle unit Design a new series elastic actuator (4th generation)
	completed		Series elastic actuators: artificial muscle profiling Roboy’s actuators are series elastic actuators developed fully in house by our team. In the scope of this project an extensive testing and profiling will be done including: frequency response (Bode plots) maximum performance range control loop update frequency on: motor driver board (FPGA) central motor control unit (FPGA) ROS layer (C++) control accuracy durability tests	For example, see https://docs.hebi.us/hardware.html#x-series-actuators
Control	LOW		Cooperative control of the mobile platform for Roboy Avatar While navigating Roboy Avatar through telepresence, it is often challenging to drive though narrow spaces or sharp corners. The field of view of the operator is limited compared to human eyes and the distances to objects are distorted. In addition, network latencies or outages might cause unsafe situations while driving the avatar. Therefore, a controller that would assist the operator with obstacle avoidance and ensure safety is desired. This project encompasses integration of the new sensors into the mobile platform and implementation of the assisted driving control algorithms.
	LOW		Assisted grasping Due to the occlusion of vision and low robot hand dexterity during telepresence, it is challenging for the operator to effectively manipulate objects. This project aims to address this issue by adding new sensing modalities to the hand and writing cooperative controllers to assist the operator during grasping and in-hand manipulation.
	MEDIUM		ROS2 motor control Migration the current motor control stack from ROS1 to ROS2 for its real-time capabilities.