Mini RHex

OSR Final Research Project

A modification to the RHex platform for swimming, walking, and climbing with the same set of legs

Robotic Locomotion and Biomimicry

Since the dawn of robotics, automated systems have been forced to operate in more and more inhospitable environments. As robotic sensing and processing has become more complex, locomotion systems have remained quite the same. The Roomba vacuum is a case in point. Roombas are an incredibly advanced robotics platform with 3D room mapping capability and object detection. However, they still only move around on two plastic wheels that struggle to change from carpet to hardwood. The motion system of one of these robots is designed to work in a wide range of environments, and as a result has several inefficiencies present in its design.

Biomimicry (the act of copying the natural world for use in technology) represents one possible approach for resolving robots' difficulty in transitioning between terrains. For instance, a robot that needs to work in the ocean without kicking up sediment may mimic the design of a stingray in order to glide through the water without disturbing anything around it. The RHex (Saranli, 2001), or Robotic Hexapod, is one such robot. The platform utilizes a tripod gait in order to move in unfamiliar circumstances. Similar to a cockroach, this gait leaves three legs on the ground in a tripod pattern while the other three can search for their next landing position. This cycle then repeats inversely as the legs previously on the ground are now searching for a new position (Saranli, 2001). By varying the speed each leg moves relative to the other two in its group, the RHex can steer or orient itself in an environment.

From Left to Right, X-RHex (Galloway, 2010), T-RHex (Martone, 2019), AQUA (Georgiades, 2004) pictured here with its Ninja Legs to allow for both land and water maneuvering (Dey, 2013)

While the RHex may possess incredible skill at walking on uneven or loose surfaces (Roberts, 2014) It can not function in a wet environment where it may need to swim short distances. And it can not climb a vertical surface with such flimsy legs. For the purposes of analysis, the locomotion approaches of the RHex can be described as walking (moving across a dry substrate), swimming (moving through water while floating or submerged) or climbing (propelling itself up a vertical or nearly vertical surface). Researchers have demonstrated that any two of these movement profiles can be combined in an RHex leg without sacrificing more than a marginal amount of performance. For example, the AQUA version of RHex is capable of walking or swimming (Georgiades, 2004), especially when utilizing its “Ninja Legs” (Dey, 2013). The T-RHex is able to climb and walk due to microspines embedded in the tips of its feet (Martone, 2019). However, no one has created a version of the RHex that can walk on land, swim in the water, and climb near vertical or vertical surfaces without necessitating a switch of the leg system. The creation of such a system would allow for the use of the RHex in extended deployment situations and mature the robot to the point of being able to complete advanced research and reconnaissance tasks in the wild.

The goal of my study was to design, fabricate and test the effectiveness of a novel RHex leg that is able to swim, run and climb. Specifically, the leg utilized the unique rounded paddle shape present in the AQUA’s ninja legs, the microspine tipped legs of the T-RHex, and the gait control system present on the X-RHex and Mini-RHex.

Methodology

Test Platform

The Mini-RHex (Barragan, 2018) is a scaled down version of the X-RHex (Galloway, 2010) platform that was used as a standin for a larger robot in this project. This robot was selected as the test platform for this project because it is small and easy to iterate upon. The platform was built as specified on its GitHub page so that results would remain consistent between units. The stock legs of the Mini-RHex are roughly the same size as my design, though mine do not compress as much as the Mini-RHex 3D printed and Plastidipped legs. In addition, my leg design is significantly heavier and thicker in order to provide a mounting location for the microspine system.

Design Theory

This project was designed with low cost fabrication in mind. The materials to fabricate all six legs can be procured for less than $100 and can be worked with using consumer grade tools. As I did not have access to a large scale machine shop to produce parts for this project, all components are designed to be as easy to fabricate as possible. The original design of the microspine system would have required continuous metal 3D printing but due to COVID-19 pandemic induced supply shortages, it was difficult to procure time on such a machine. As such, the system was redesigned in order to use fewer parts that could be fabricated using equipment in my workshop.

Leg Design

The novel leg design illustrated in this project incorporates aspects of the legs present on the X-RHex, T-RHex, and AQUA robots. The leg is a large offset cam that “Walks” by impacting the ground at a specific point in its rotation. On the leading edge of the leg is a set of microspine cartridges that are used while climbing. Within the mounting bracket, each microspine cartridge is able to rotate freely to mesh into the object being climbed. Inside the cam is a flexible plate used while swimming as a paddle.

Sub-components present in the Novel Leg. Side Plates (1), Spacer Plate (2), Shaft Connector (3), Rear Spacer (4), M3 Bolt (5), M5 Locknut (6), Microspine Cartridge (7), Microspine Cartridge Bracket (8), Swim Plate (9), Retaining Flange (10)

Microspines

The microspine subsystem utilizes a set of two fish hooks per leg encased in a 3D printed cartridge. Similar to Martone (2019)’s approach, the hooks are inserted during a 3D print. A void is modeled into the design of the cartridge for the hook to be placed and partway through the print, the hook is inserted into this void. Note that this method of fabrication requires an operator to be present throughout the 3D printing process and as such may not be feasible if these legs needed to be produced en masse.

By manufacturing the micrispines inside cartridges, the hooks become a modular unit that is much easier to install and maintain inside the larger Mini-RHex system. In case of damage to a single microspine cartridge during a climbing operation, each cartridge is designed to be field replaceable utilizing only a single hex key.

A diagram of the microspine cartridge. 2 cartridges were attached to each front leg, for a total of 4 spines.

Swim Plate

Inside the leg is a flexible plate of acrylic used as a paddle when swimming. This element was modeled off of the “Ninja Legs” for the AQUA designed and demonstrated by Dey (2013). The paddle is held in place at the top of the leg but is able to swing freely at the other end, allowing the paddle to bend and flex without breaking as it moves. Several iterations of this design have been created, including several that utilize a “live hinge” mechanism to allow for further flexibility in the plate. Ultimately, a rigid design was selected because the plate could also be utilized in order to hold up the bottom of the leg during walking operations.

A side view of the leg assembly, the swim plate is highlighted in red

Leg

The leg itself is constructed from laminated acrylic cut on a laser cutter. The three layer design leaves a gap in the center of the leg for water to flow freely through. A 3D printed stabilizer is present toward the rear of the leg to prevent the panel from splaying apart while walking. During preliminary tests, the legs proved to be too slippery across a multitude of materials. To resolve this, dabs of rubberized plastic were added to the bottom of the leg to increase traction.

The final novel leg design render

Test Procedure

The robot was placed at a marked starting location and was tasked with moving in a straight line to the end location 3 Meters away. During the test, the number of times a leg “slipped” or failed to make proper contact with the ground was catalogued and the time it took for the robot to reach the finish line was logged. The walking tests were conducted with both the stock Mini-RHex legs and the Novel Leg concept. Each test was repeated 10 times.

During hang testing, the Mini-RHex was placed on the side of the tree and hung unassisted for 30 seconds. The hang test was repeated 10 times and at the end of each 30 second window, the total number of hooks engaged was catalogued. The gait to transition from walking to climbing is currently in development and as such is not tested here.

COVID-19 related issues with obtaining the proper controller system prevented in-water testing, and further testing is required to fully analyze the effectiveness of the novel leg design while swimming.

Results

When the Mini-RHex was started with the new prototype legs attached, the platform began to shake violently after approximately 15 seconds. To counteract this, a round of Proportional Integral Derivative tuning was applied that adjusted for the higher leg weight. In the future, the new PID table will be uploaded to an otherwise unchanged codebase when switching to the novel leg design.

Land Testing

The Novel legs averaged a trial time of 23.4 seconds with a shortest run of 20 seconds and a longest run of 35 seconds. During its trials, the novel legs slipped an average of 0.6 times per trial though five of the trials had 0 slips, four trials had 1 slip, and only one trial had 2 slips. The novel leg failed to complete the test in trial one due to PID induced shaking and thus is not factored in these calculations.

The Control legs averaged 21.4 seconds per trial with a shortest run of 20 seconds and a maximum of 23 seconds. The control legs slipped an average of 0.5 times per trial, though five trials had zero slips and 5 only had one slip.

Hang Testing

Over the course of 10 trials, a catastrophic failure only occurred on trial 9 in which a piece of bark peeled away from the tree, taking most of the robot with it. The average hook engagement was 3.5 hooks per trial.

Water Testing

Due to COVID-19 related setbacks, it was difficult to procure a waterproofed version of the OpenCM-9.04 board in time for testing. Water testing of this leg could be a topic for further research.

Discussion

The goal of this study was to combine aspects of the X-RHex, AQUA, and T-RHex legs in order to switch from a walking gait to a swimming or climbing gait without altering any components on the robot. The results of this trial are promising however more research is needed in order to further smooth out the transitions between gaits and further increase the accuracy of the legs so that potentiometer based position tracking could be implemented in the future.

While the Mini-RHex was selected in order to develop the novel leg design without necessitating investment in a full-sized RHex system, this design could be easily adapted and scaled to fit the larger X-RHex platform. Another way to move forward could be the merging of the novel leg design with the more complex adjustable spring tension system outlined by Galloway (2013). Such a design would allow for the introduction of long distance leaping and more efficient galloping and pronking (springing into the air) when moving on land and the possibility of deploying a much more efficient in-water propulsion system. The novel legs outlined in this project may also benefit from a tail as seen on the T-RHex to allow for safer and more efficient climbing.

The potential use cases for a truly all-terrain robot are endless. Researchers looking to study coastal conditions both on land and in the water could utilize such a robot to autonomously gather data over an extended period. Search and Rescue teams could utilize the Mini-RHex with a novel leg system to search through wreckage after a tsunami where the robot may need to climb out of bodies of water onto rubble. And climate scientists could deploy sensor packages deep into forests for months at a time to gather long term climate data.

Limitations

The majority of the time in the study was dedicated to the development of hardware components and PID tuning. As such, the gait controller does not contain all the necessary transitions between each gait and may become stuck or fail to properly transition. To fully test the potential applications of an all terrain RHex, another study could be conducted that focuses more on gait development and testing than hardware design. In addition, it was difficult to procure a waterproofed version of the OpenCM-9.04 board in time for testing. In-water component of my study was conducted using a computer model instead of tank or open-water testing.

While the results of this research appear promising on the Mini-RHex, testing on a full sized X-RHex is necessary in order to fully evaluate the design.

References

Saranli, U., Buehler, M., & Koditschek, D. E. (2001). RHex: A simple and highly mobile hexapod robot. , (7), 616-631.
Galloway, K. C., Haynes, G. C., Ilhan, B. D., Johnson, A. M., Knopf, R., Lynch, G. A., ... & Koditschek, D. E. (2010). X-RHex: A highly mobile hexapedal robot for sensorimotor tasks.
Roberts, S. F., Duperret, J. M., Li, X., Wang, H., & Koditschek, D. E. (2014). Desert RHex technical report: Tengger desert trip.
Roberts, S., Duperret, J., Johnson, A. M., van Pelt, S., Zobeck, T., Lancaster, N., & Koditschek, D. E. (2014). Desert RHex technical report: Jornada and white sands trip.
Georgiades, C., German, A., Hogue, A., Liu, H., Prahacs, C., Ripsman, A., ... & Milios, E. (2004, September). AQUA: an aquatic walking robot. In (Vol. 4, pp. 3525-3531). IEEE.
Dey, B. B., Manjanna, S., & Dudek, G. (2013, November). Ninja legs: Amphibious one degree of freedom robotic legs. In (pp. 5622-5628). IEEE.
Martone, M., Pavlov, C., Zeloof, A., Bahl, V., & Johnson, A. M. (2019). Enhancing the Vertical Mobility of a Robot Hexapod Using Microspines. .
M. Barragan, N. Flowers, and A. M. Johnson. "MiniRHex: A Small, Open-source, Fully Programmable Walking Hexapod". In Robotics: Science and Systems Workshop on ``Design and Control of Small Legged Robots'', Pittsburgh, PA, June 2018.
Galloway, K. C., Clark, J. E., & Koditschek, D. E. (2013). Variable stiffness legs for robust, efficient, and stable dynamic running. , (1).
Moon, H., Cho, J., & Kong, K. (2018, October). Retractable cleat mechanism of legged robots' foot on various terrain. In (pp. 918-921). IEEE.

All Done?

Back to Projects