Modboats

Motivation

Modular robots offer a pretty amazing promise: build simple generalist robotic modules that can move and connect together, and you can turn those modules into virtually anything. Roboticists are working on this vision for land robots, but it’s pretty far away for aquatic robotics. This is largely due to the challenge of getting precise and holonomic motion in water, which has historically required lots of motors and complex moving parts. This means the modules are expensive, and not many get built. We aim to disrupt the paradigm that such motion is necessary, by building an underactuated, inexpensive aquatic robot that can be a modular self-reconfigurable robotic system (MSRR).

The Modboat is a single-actuated DOF aquatic robot designed for low-cost surface operations and ready for modular operation. It consists of two concentric cylinders connected by a motor. The motor spins the upper body relative to the lower one, which causes the lower body to counter-rotate (to conserve angular momentum). This angular acceleration opens one of two passive flaps on the lower body, which then pushes the water and produces forward thrust. By rotating in a sinusoidal fashion, the Modboat can oscillate around a straight-line trajectory; by cleverly manipulating the rotation rates, the Modboat can also turn using the same strategy. This design is based heavily on work by Refael and Degani [1][2].

This motion is not easy to control precisely, but this coupling of rotation and translation allows the Modboat to move in two dimensions using only a single actuator. This makes it inexpensive, which means it can be deployed in swarms over large areas or to conduct ocean research, clean oil spills, or assemble into ocean infrastructure.

Design Work

Initial work on the Modboat focused on meeting the following two design criteria:

1. Create a simple design that can be produced from stock materials.

2. Modify the original design by Refael and Degani to make it ready for modular self-assembly.

The design is shown in the photo above. The shell is built from acrylic tubing and sheet stock, which can be easily joined with a waterproof seal. The interior design is laser-cut ABS, which holds the motor, electronics, and batteries in place. Although this design was successful and served well for years, we found that the 6” OD acrylic cylinder was difficult to source and would often deform when shipped, so the newest Modboats feature a single-piece top body that is 3D printed using SLA, which forms a watertight print and incorporates all of the mounting features.

Unfortunately, we found that the initial design suffered from a very high sensitivity to symmetry, causing it to sharply turn depending on how it was balanced. We were able to identify the source of the sensitivity - which was the way the flippers amplified any slight imbalance - and correct it. This work was presented at the 2020 International Conference on Ubiquitous Robots (UR) [3], and the virtual presentation I gave is linked below.

Control and steering

In order to build a modular self-reconfigurable system, it is first critical that the modules be able to move. As mentioned before, the Modboats’ propulsion method is not easy to control, so the first challenge was to build effective control for the system. Between 2019 and 2021 I developed three ways of controlling the Modboat:

1. Differential Thrust

2. Inertial Steering

3. Desaturated Thrust Direction (DTD)

Differential Thrust steers the Modboat by modulating the frequency of the motor oscillation. Spinning faster one way produces more thrust, so the Modboat moves forward and turns. While Refael and Degani used this to great success [1], I found that it suffers from low control authority and so is not an effective control method.

Inertial Steering steers the Modboat based on drift. By pausing the motor at the peaks of rotation, we allow the Modboat’s natural drift to turn it in between cycles. This strategy has high control authority, but is nearly impossible to plan for when coordinating multiple modules, because time is an input to the system [4].

Desaturated Thrust Direction (DTD) provides the best of both worlds - high control authority and easy planning. By applying a control law formulated to treat the Modboat bottom body as a forced pendulum, we can cause the Modboat to oscillate around the stable equilibrium, which is a function of a reference heading. Since oscillations create the paddling motion the Modboat uses to propel itself, this means that it can function as a pointable thruster. In work presented at IROS 2021 [6], we have shown that - under this control law - the Modboat is significantly more maneuverable than any prior control method. Even with only one motor, it can execute a 180 degree turn in under one body-length, and follow trajectories only 3.3 body lengths long.

Docking and Undocking

Now that each module can move effectively in the water, the next step to building an MSRR is docking and undocking from other modules. To achieve that, we added four magnetic docking points around the circumference of the top body, and designed an algorithm for docking. This is particularly challenging because all parts of the Modboat are turning, and we can only control its orientation on average. So how can we dock - an operation that requires precise orientation control?

The key is to separate the process into manageable chunks. One Modboat is the (stationary) target - sitting and doing nothing, while another approaches to make the connection. Initially, the algorithm calls on the moving Modboat to get some distance, and then approach the desired dock on the target along a straight line. The dock is aborted if the approach isn’t straight enough, but if not then the moving Modboat can establish a drift velocity that can take it to the target. At that point it can stop swimming and transition to controlling its orientation precisely, which it does by using the bottom body as a reaction wheel for the top body. Using this method we can successfully dock in either a forward- or side-facing orientation over 95% of the time.

How about undocking? Can the Modboats separate once they’ve connected? Yes they can! The Modboat has a tail on the bottom body that protrudes from the top body footprint. If two neighboring boats bring their tails together quickly, they can separate and drift apart. If the motion isn’t quick enough, they will redock in a different configuration, which we can use as a feature to reconfigure the docked boats. This work has been accepted for publication at the 2021 IEEE International Conference on Robotics and Automation (ICRA) [5]

Future work

In future work we hope to consider effects of external flows on the robot. We also intend to explore the effect of modularity on the system to determine whether multiple Modboats connected together can improve on their individual swimming abilities.

References

[1] G. Refael and A. Degani, “A Single-Actuated Swimming Robot: Design, Modelling, and Experiments,” in Journal of Intelligent and Robotic Systems: Theory and Applications, 94, 471–489. https://doi.org/10.1007/s10846-018-0776-x.

[2] Gilad Refael and Amir Degani. “Momentum-driven single-actuated swimming robot.” 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Hamburg, Germany, pp. 2285-2290. https://doi.org/10.1109/IROS.2015.7353684.

Publications

[3] G. Knizhnik and M. Yim, “Design and Experiments with a Low-Cost Single-Motor Modular Aquatic Robot", 2020 17th International Conference on Ubiquitous Robots (UR), Kyoto, Japan, pp 233-240, doi: 10.1109/UR49135.2020.9144872.

[4] G. Knizhnik, P. deZonia, and M. Yim, “Pauses Provide Effective Control for an Underactuated Oscillating Swimming Robot,” in IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 5075-5080, Oct. 2020, doi: 10.1109/LRA.2020.3005383.

[5] G. Knizhnik and M. Yim, “Docking and Undocking a Modular Under-actuated Oscillating Swimming Robot,” 2021 IEEE International Conference on Robotics and Automation (ICRA). Xi’an, China. In Press. pp 6754-6760, doi: 10.1109/ICRA48506.2021.9562033.

[6] G. Knizhnik and M. Yim, “Thrust Direction Control of an Underactuated Oscillating Swimming Robot,” 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Prague, Czech Republic (Virtual). pp 8665–8670, doi: 10.1109/IROS51168.2021.9636778.

[7] G. Knizhnik, P. Li, X. Yu and M. A. Hsieh, "Flow-Based Control of Marine Robots in Gyre-Like Environments," 2022 International Conference on Robotics and Automation (ICRA), Philadelphia, PA. pp. 3047-3053, doi: 10.1109/ICRA46639.2022.9812331.

[8] G. Knizhnik and M. Yim, “Amplitude Control for Parallel Lattices of Docked Modboats”, 2022 IEEE International Conference on Robotics and Automation (ICRA). Philadelphia, PA. In Press. pp. 3027-3033, doi: 10.1109/ICRA46639.2022.9812381.

Modboats have been featured!

Check out these articles about the Modboats:

1. TechXplore: “Modboat: A low-cost aquatic robot with a single motor”

2. Hackster.io: "Modboat Is a Low-Cost Robot That Swims Using Only One Motor"

3. Modlab project feature: Modboat: A Single-Motor Modular Self-Reconfigurable Robot

Video Presentations

I’ve given a number of presentations on my work with Modboats. There’s a playlist on Youtube, and they’re linked below:

1. “Design and Experiments with a Low-Cost Single-Motor Modular Aquatic Robot" - presented at the 2020 17th International Conference on Ubiquitous Robots (UR).

2. “Pauses Provide Effective Control for an Underactuated Oscillating Swimming Robot” - presented at the 2020 IEEE/RSJ Conference on Intelligent Robots and Systems (IROS).

3. “Docking and Undocking a Modular Underactuated Oscillating Swimming Robot” - presented at the 2021 IEEE International Conference on Robotics and Automation (ICRA).

4. “Thrust Direction Control of an Underactuated Oscillating Swimming Robot” - presented at the 2021 IEEE/RSJ Conference on Intelligent Robots and Systems (IROS).