Tensegrity Robotics

One area of exploration in robotics that is getting a lot of attention recently is building robots using tensegrity principles. From the NASA team who pioneered the tumbleweed exploratory robot known as the Superball Bot https://www.nasa.gov/content/super-ball-bot to the intriguing duct robot developed by Jeffrey Friesen at UC San Diego http://www.sunspiral.org/vytas/cv/DuCTT_ICRA_2016.pdf, innovative methods are being developed that attempt to utilize tensegrity to build highly unusual robots. There are certain advantages to employing tensegrity principles but some real difficulties and disadvantages too. In a paper I wrote with my colleague Dr. Dorothea Blostein from Queens University, we explore some new approaches to mechanizing tensegrity structures. The paper was presented at the 16th Biennial ASCE conference held in Cleveland April 9-12, 2018 and you can find it here.

More specifically though, are  a myriad of technical problems that need addressing before the best approach to  tensegrity construction is known. These can be broken down into basic categories as follows:

  1. Internal actuation of a tensegrity mechanism. If the internal approach is used this requires loosening and tightening the tension net within the tensegrity to control it’s prestress and hence it’s ability to move. By altering the relative tension in various tension lines or cables the shape of the tensegrity changes. The most successful application of this technique is probably the Superball Bot built by Vytas Sunspiral’s team at NASA. By altering the shape of a simple 6 strut expanded octahedron tensegrity they were able to cause it roll as it’s centre of balance continued to change. While I am told that they are able to recover some of the potential energy expended when the lines change tension it is my hunch that altering the prestress in this way is a net sum energy loss. This may not matter if there is sufficient energy available to the system but if not there may be better ways to move a tensegrity around. Additionally, their prototype so far has no means to propel itself out of a large enough hole or up a hill.
  2. External activation of a tensegrity complex. As the above paper details there may be a better way to employ tensegrity which involves attaching fixed tensegrity modules to each other by means of saddle slings or other forms of connection .This creates revolute and prismatic hinges of various sorts which can then be controlled separately by a third set of actuator lines. This includes the possibility of pseudopods or extruded sections that  create contingent legs that could push against the environment and allow it to walk or pull itself out of trouble.
  3. Hybrid Structures. It’s tempting to employ traditional bio-mechanical solutions to the tensegrity system but in general I think it is a mistake. Tensegrities are so unlike all other kinds of manmade structures that entirely new principles and methods need to be worked out from first principles. While some form of hybrid may work it is the interface between the tensegrity part and what ever it attaches to that becomes problematic. The reason to use tensegrity components is to reduce weight and increase resilience – under sufficiently high prestress a tensegrity mast can serve as a reasonably rigid appendage though there will always be some amount of compliance built in. The approach of traditional bipedal and quadrupedal robots such as those created by Boston Dynamics is to essentially attach legs and arms to a chassis using multiple degrees of freedom ball and socket joints with cable driven or hydraulic actuators. They have been able to accomplish amazing feats of engineering and combined with machine learning algorithms have created  robots that can perform many tasks that even a few years ago would have be seen as impossible. However, traditional mechanical solutions have limitations that a tensegrity may solve. They tend to be heavy structures that require a lot of energy and more importantly traditional joints are vulnerable to damage and impairment that may not be fixable in the field. In contrast tensegrity joints are really disjoints, that is, they are formed by relaxing the triangulation somewhere in the structure to allow for movement. It’s possible that such joints can be designed to disperse loads through multiple vectors and thus a damaged joint could feasibly be workable albeit in a reduced capacity.
  4. Tensegrity Joints  There are a number of ways that a ‘disjoint’ can be created in a tensegrity. In a closely coupled helical tensegrity mast rhombic facets are created owing to the weaving pattern and these can be manipulated to cause the mast to contort much like a snake. The same motion can be duplicated using a series of stacked stellated tetrahedrons linked with saddle slings. This will look much more like a series of vertebrae and there are several ways that actuator lines can be used to control such a mast. But linking discrete modules together, for example a series of tensegrity prisms to create an operational leg with something like a knee joint, involves a mechanism that looks on the surface to be very much like a typical revolute joint. Compression components rest on tension components and swivel around them so issues of chafing, wear and friction are going to be prevalent just as in normal joints. But a tensegrity joint is different in that the hinge mechanism is not  fixed and rigidly constrained which may create unique problems of it’s own. I think these are engineering problems that can be solved but there is lots of research needed here.
  5. Assembly and rapid prototyping  Tensegrities are hard to build and hard to balance. This factor alone has impeded comprehensive development of all the design perimeters possible. In the last few years computer scientists have taken over and created very useful and functional software programs that allow form finding and actuation algorithms to be tested in simulated environments. Still, eventually prototypes will have to be built and best methods for assembly fine tuned. Over the years my system of employing elastic cords and slotted dowels has allowed me to build hundreds if not thousands of tensegrities rapidly and robustly. I can test out my ideas quickly and make changes on the fly. While computer simulations have mostly supplanted this method of exploration I still think it will have value when transitioning designs into the real world. I plan to create a number of instructional videos in the next while and post them on my youtube channel and on my site showing my techniques in detail.