Figure 2. The three-segment MSMR will be used to develop coordinated control software for the system.
The linkage, which makes the mechanical connection between the robot modules, must be flexible to allow the robot to turn and maintain wheel contact with the ferrous surface being climbed. It must also be able to transfer push (compressive) and pull (tension) forces between the robot modules so they can work in concert to overcome obstacles greater than the capability of any one robot module.

Drive modules with too many degrees of freedom (DoFs) may make control of the system overly complex and limit how force can be transmitted between robot modules. On the other hand, a linkage with insufficient DoFs will limit the maneuverability and obstacle traversing ability of the system. To better understand the linkage needs, a series of prototypes was designed, fabricated, and tested.

The elastomer bow linkage (EBL) was pursued because it appeared that it would provide both flexibility for maintaining wheel contact and the transmission of forces for push-pull benefits. At rest, the bow linkage is rainbow-shaped, allowing the robot to traverse external corners more easily. The design is extremely simple, robust, and easy to fabricate. Linkages from three different durometers of elastomer were fabricated and tested. Testing results showed the elastomer bow was effective at translating tension forces and providing flexibility to maintain wheel contact. However, during some cases when push force was needed, the elastomer linkage started to buckle and the rear wheels caught up to the front, creating a situation where the robot stuck to itself. For this reason, the hardest durometer linkage worked best but still had problems with buckling. Implementing a harder durometer elastomer would limit turning flexibility and would not allow for adequate pitch flexibility on a system with three or more robot modules (internal robot modules would lose surface contact while traversing internal corners).

It is conceivable the implementation of the coordinated control system may reduce the cases where the linkage is buckled. However, the sensors required to maintain information on relative position of robot modules are much more complex than linkages with discrete DOFs. It is expected that some type of vision or beacon system would be required to implement a coordinated controller for a system with this type of linkage.

A rigid-bow linkage was fabricated for evaluation on the two-segment prototype robot. The linkage looked exactly the same as the EBL, except that it was fabricated from a piece of plywood that made the entire robot structure rigid. The robot could traverse obstacles similar to those of the EBL as long as it was perpendicular to the obstacle. The rigid linkage did not perform well if the obstacles were approached at an angle, because not all four wheels could maintain contact with the surface being climbed. The rigid linkage also provided very poor turning capability.

A linkage design that includes a three-DOF ball joint that can lock the pitch-and-yaw DOFs when compressed was designed, prototyped, and tested. The idea is that for compressive force to be translated from one robot module to the next, the ball joint must be locked in the pitch DOF or the linkage will simply start to fold on itself. The ball-joint design allowed the wheels to maintain surface contact with pitch, roll, and yaw DOFs. The design also allowed pull forces through the ball joint. Inside the ball joint, a spring pushes the ball into the normal operating position, but when a compressive force between the robot modules starts to compress the spring, the ball slides further into its receptacle, and the pitch-and-yaw DOFs are constrained. As the compressive force diminishes, the spring returns to its original position and the joint regains its full range of motion.

The spring implemented on this design was too rigid and unable to be compressed by the robot modules. Additionally, the range of motion of the ball joint was not enough to allow the robot to traverse external corners. Conceivably, the lockable ball-joint design could be implemented on a bow-shaped linkage, allowing the traversal of external corners. A spring with a lower spring rate could also be integrated into the system to better evaluate the feasibility of this design.

For a robot with only two drive modules, a pitch DoF is not needed to maintain wheel contact when traversing an internal corner. For this reason, a single DoF bow linkage was developed. The first prototype simply included a hinge point in the middle of the bow linkage. The linkage proved effective for turning, but if the system was accidentally turned too much (which happened often with the manually controlled prototype), wheels from the adjacent robot modules could get stuck together.

To mitigate this failure mode, a second prototype was developed with a limited range of motion at the hinge point. Instead of using a traditional hinge, a leaf spring was designed as the pivot to try to bias the robot towards a position where the front and rear wheels were aligned. The design resembled a simplified saloon door concept. The wires connecting the front and rear robot module were run through the middle of the linkage and along the sides of the leaf spring. The pivot and range limiting portions of the design worked well, allowing effective turning without allowing wheel-to-wheel contact. The spring contribution of the design was ineffectual since the torque produced by the wheels could easily overpower the spring. For future designs, using a spring to provide a bias at a joint is ineffective for this application. The coordinated control system can implement the robot module relative position control more effectively and efficiently.