• underwater robotics
• mechanism design
• docking mechanisms
• legged robots
• modular robotics

Underwater robotics has many current and potential applications, including environmental monitoring and cleaning of the sea, marine exploration and research, maintenance of underwater structures and defense. The underwater domain is difficult to negotiate, even for robots, and poses challenges in mechanical design, control, instrumentation and communications. Underwater missions often call for long-term deployment and sometimes navigation over large areas. Certain applications may involve transitioning among multiple locomotion modalities; for example, robots might have to swim in the pelagic zone and also move in close contact with the seabed (benthic zone), navigate on the water surface, shallow water or even terrestrial environments. Underwater manipulation is a difficult task, too, since it usually requires active control of many degrees of freedom for stabilization against the disturbances of marine currents.

Any robotic self-contained unit with the versatility to do most of the aforementioned tasks is bound to face issues of cost, power and time efficiency or hardware and software complexity. On the contrary, cooperative systems and, in particular, modular swarm robotics take advantage of the physical and cognitive interactions between relatively simpler and non-specialized units to execute complex and varied activities. Swarm robotics is, indeed, a growing field in the underwater robotics community for this potential benefits.

In this work, it is argued that underwater legged robotics, a novel research area that addresses the limitations of crawling robots, can take off by exploiting modularity and swarm behaviour. Even though legged robots have proven to perform well on precise surveying, disturbance rejection, manipulation and environmental impact on the seabed, their practical applicability is being held back by their restricted capacity to navigate the environment and their complex mechanical architecture which makes them prone to failure. Modular swarms of legged robots not only could improve performance in the benthic zone (e.g. time efficient concurrent search, fault-tolerance, self-reconfiguration for manipulation tasks, negotiating obstacles and adaptive locomotion) but can also enable pelagic navigation by cooperation with swimming robots.

Taking into account that one key factor for scalability and versatility of multi-robot systems is the simplicity of their robotic units, the case of one-legged robots is worthy of studying. Underwater single-legged robots use punting locomotion, a type of gait where the leg propels the unit upwards and forward by pushing against the ground to take off for a ballistic gliding phase. They are statically unstable but can be devised to work with limited sensing feedback and no other actuators if they have a self-stabilizing morphology. From a modular robotics perspective, they constitute the building block for any other configuration of multi-legged robots.

Modular robots physically connect one another by a process known as docking in order to form more complex morphologies that increase their capabilities. So far, the scientific literature on modular robotics has only covered mechanisms and control strategies for static docking, namely, where the modules can align themselves with closed-loop control to ensure viability of the coupling procedure. Statically unstable robots with minimal closed-loop feedback, require different approaches to the docking problem (dynamic docking).

In this work, a robotic dynamic docking system was proposed for a one-legged robot that replicates the bio-inspired underwater inverted pendulum model (USLIP) for punting locomotion. The mechanism is based on a spring-loaded gripper actuated by a single degree of freedom and triggered by mechanical buttons that sense collisions with the target. Alignment sensing and control is handled via morphological computation. The robotic leg has two degrees of freedom: one for the hip and one for pushing. The control and sensing of both the leg locomotion and docking system are also discussed here. A prototype of the leg with the docking system, was manufactured mostly with rapid prototyping techniques and the control was implemented in a development board.

The robot, tethered to an external power supply and controller, was experimentally tested in a water tank. Its positions over time were estimated using a camera-based two-dimensional motion capture system. The kinematics of the robot was described in terms of expected value and variance of significant locomotion variables. A passive target module was used for testing the docking effectiveness.