Robotics

The Centre for Biorobotics studies how robots can move in a variety of environments, from fully submerged swimming to walking on land, and creates locomotion systems and their control methods, often taking inspiration from animals who have adjusted for life in these conditions. These novel actuation mechanisms allow efficient, robust, or versatile locomotion. 

Control

Underwater robots can move in 6 degrees of freedom in environments with disturbances and rely on control strategies to plan and execute movements and tasks. A common disturbance source is water motion, which can cause a robot to deviate from its planned path. Also, when the robot is moving, the water around it is moving too in quite unpredictable ways. We study how control algorithms can generate effective locomotion and address these challenges.

Aquatic animals use a lateral line sensing organ to detect hydrodynamic events to aid their locomotion and navigation. The FILOSE robot uses a pressure-based Artificial Lateral Line sensor array to control its position with respect to the flow. It can find spots in turbulent flows where drag is smaller and hover there to save energy.

The U-CAT robot is very maneuverable and versatile, which makes it suitable for underwater applications where high maneuverability is needed. A challenge for U-CAT is executing tasks in turbid waters with low visibility. Our work on driver tracking uses a combination of vision and radio signals, and a sensor fusion algorithm that finds the best combination of the two sensing modalities. When the diver is out of sight, the controller relies on radio signals to locate the diver, and when the diver is visible, these visual cues are more trusted. 

Robot control can also be implemented using Machine Learning. Instead of deriving mathematical models for control algorithms, learning algorithms can be used to teach the robot to move in a desired way. Machine learning is especially suitable for robots such as U-CAT: an underactuated robot with delayed and coupled response, which interaction with water is difficult to describe in equations.   

Bio-inspired Actuation

FILOSE is a fish-inspired robot prototype that moves using a soft tail. We have shown that by mimicking the material properties of fish, an actuation pattern similar to real swimming fish can be created without complicated distributed actuation, using only one motor. Additionally, asymmetric motion patterns of the tail can be used for control of the robot's orientation.

U-CAT is a concept-vehicle inspired by turtle locomotion. It has four independently driven flippers which allow motion in 6 degrees of freedom. U-CAT was originally built for underwater archaeological and shipwreck surveys. It is highly agile, allowing moving in tight spaces and following contours of underwater objects. The flippers do not create strong water jets and therefore they do not stir up sediments from the sea bottom, making visual observation possible. U-CAT has been used for investigating archeological sites, underwater vegetation and fishfarms. 

μ-CAT (micro-CAT) is a small version of U-CAT which serves as a platform for research, student projects, and public outreach. Most fin-actuated robots rely on lift-based forces for propulsion. However, drag forces, normal to the surface of the fin, can be overpowering, but are usually considered parasitic. We designed a bird-inspired fin that allows us to use the drag forces for propulsion.

Locomotion on soft grounds (such as mud, snow or wet sand) is challenging for robots. Moving in viscous of frictional granular media is very energy consuming, and the properties of the matter are varying and more challenging to predict than those of the hard ground. Therefore, it is also hard to describe and predict the interaction between the robot and the wet matter. We investigate both mechanical solutions and control algorithms that make it possible to use robots on those difficult natural terrains.

Legged locomotion

Many animals navigate muddy environments, inspiring our study of biomechanics and deformable terrain dynamics. We aim to understand the intricate interactions between soil and appendages to enhance control systems and optimize foot designs for quadrupedal robots. Through this research, we seek to simulate and replicate the features which enable these animals move through such challenging landscapes. Our goal is to engineer more efficient and effective robotic locomotion in muddy terrains.

Archimedean screw-based locomotion

The ROBOMINERS project aims to reduce the environmental impact of mining and the risk to human operators of machinery, by developing automated solutions for selective mining. The small-scale RM3 prototype allows us to study locomotion, perception, and navigation concepts. Its locomotion system uses four individually controlled Archimedean screws, that can generate holonomic motion control. Locomotion control strategies have been tested in a variety of terrains.

Traditional methods of locomotion perform sub-optimally in unstructured and unpredictable environments, as well as when transitioning between aquatic and terrestrial conditions. We proposed a reconfigurable, soft fluidic actuator that can dynamically modify its shape and stiffness to enable locomotion in terrestrial, aquatic, and multiphase environments.