Mechanical Overview

The robot uses eight legs, two on the right and two on the left for steering, and four in the back for balance. It uses a differential drive system to steer the robot in the direction specified by the camera. We went through three iterations of the robot shown below. The first was a prototype primarily used to test the legs and develop an understanding of the problems we would encounter. The second iteration took steps towards developing the form of the beest, but some design problems prevented it from fully working. The third iteration improved on the second to produce a functional beest.

Leg Design

Overall Leg Design

The leg design is based on Jansen's linkage, a linkage design made by Theo Jansen for his Strandbeest sculptures. The legs needed to transfer the rotation of the motor shaft into a stepping motion. The initial design (shown right) was made to replicate the motion of the Jansen linkage. Assuming we would need to make small adjustments, the first iteration of the linkages were designed with future modification in mind; each linkage contains bolt holes at 1 cm intervals to make small changes very easy. This ended up causing strength problems within the part so after the final lengths were determined the design shifted to solid parts (shown left) in the second iteration. The other change between the two designs are the modified linkage lengths. While the initial design was optimized for the step pattern, the second iteration was designed for simplicity while keeping a step pattern that would work on the robot. This simplification led to using five 9 unit linkages making construction much easier.

Material Choice

The first iteration of the legs used 3d printed linkages made of PLA printer filament. This material was chosen because it’s a readily available prototyping material that our team had experience using. The material proved to have too much friction for the motors we were using, causing the legs to bind. In addition the print time for each leg was roughly 5 hours, and printing 8 legs for the final design would have been a challenge.

We decided to move to ⅛th inch laser cut acrylic for the second iteration. This allowed us to cut all 8 legs on one sheet, reducing the manufacturing time significantly. The acrylic also proved to have lower friction than the PLA, and when combined with the higher torque motors we purchased (this is explained under motor selection) this solved the friction problem. However, the new motors increased the weight of the robot and the acrylic was not strong enough to counter the mass increase. The legs bent, and would not function properly when holding the full robot. To solve this we increase the ⅛th inch acrylic to ¼th inch acrylic. While there's still some bend in the legs, they are functional and are unlikely to break.

Feet

Another factor that helped the bending problem was the addition of 3d printed feet. We made feet for each leg that prevented them from sliding to the side (shown Right). Initially the feet were just plastic, however on the tile floor the robot is designed to walk on, they couldn’t get enough traction to actually move the robot forward.

To solve the traction problem we added rubber bands to the pads of the feet. But the bending of the legs meant that the feet would drag while trying to move forward. The solution was to make feet that only had friction in one direction, pulling the robot forward then sliding back into position. This was accomplished by putting the rubber bands on an angle, and adding a layer of tape to the top. This means when the foot is moving forward the foot slides along the tape, and when it is moving back the angled rubber catches the ground pulling the robot forward.

Chassis

The chassis consists of 5 leg linkage mounts and a central chassis box which holds the electrical and hardware components of the robot.

Leg linkage mounts consist of 2 hardboard plates held together by PLA brackets which attach the leg linkages to the main chassis box. Leg mounts contain a space for the linkage’s fixed axle and a secondary space for the axle which powers the linkage, either directly from the motor shaft or through the offset system. 3 leg mounts hold motors, while 2 hold the offset axles of back legs in place. Leg mounts are attached to hardboard plates at the top and bottom, forming the central chassis box.

The chassis box is the main body of the robot, and holds the Arduino and motor shield, the battery, the camera, the RasPi, and a second battery for the RasPi. The motor shield, arduino, batteries and cables are all stored in spaces formed by the leg mounts to conceal them and give the robot a tidier appearance. The Raspi is secured under the main body to lower the center of mass and make the robot more stable, while the camera is mounted on top.

The chassis plate is designed with slots at the leg mounting points. These slots allow the motor mounts to be slid along the chassis plate then re-tightened for on-the-fly repositioning of the legs-motor components.

The chassis plates were made of hardboard such that holes could easily be drilled in the plate to reposition fastening points for the RasPi and camera, which are held outside the main chassis box. Hardboard is also a fairly rigid material, which (especially configured in the box shape of the chassis center) will resist bending under the weight of the motors electronics.

Motor Selection

The primary factor in choosing motors was the torque, and speed was secondary. Our original motors were rated at 1.5 kg cm of torque with a speed of 200 rpm. After encountering issues with stalling, we selected motors rated at 7 kg cm of torque, 4.5 times higher than the original motors. Though the motors were heavier and the speed was much slower at 60 rpm. The speed is not a problem for the functionality of the robot, however, because of the friction decrease in the linkages, a higher speed lower torque motor would have been sufficient. The increased mass did prove to be a problem as explained in the section on material choice for leg design.

Motor Axle Mounts

The legs all had to be driven by an offset axle because the movement of some of the linkages would interfere with a traditional axle. To address this with the front legs we chose motors that had an axle coming out of either side so the motor could be placed between the two legs to eliminate this interference. A 3D printed part was used to turn the rotation of the motor shaft into an offset axle movement to drive the legs. Each side was rotated 180 degrees from the other so the legs in the pair are in opposite positions from each other at all times. The motor shaft mount CAD is shown below as well as images of the assembled connection between the motor and front legs.

We had challenges with the motor shaft deforming the PLA allowing the motor to turn freely so we added a heat seat insert and a set screw to the part to stop this from happening.

For the set of back legs, all 4 legs are driven off of the same motor. To accommodate for this, we designed an extra part to act as an extension of the offset axle and transfer movement from the inner legs which were driven by the motor to the outer legs. It splits in half so that it can be inserted into the chassis panels that support it and has concealed teeth to lock the two sides at an offset angle. While the teeth themselves are not particularly strong in a 3D printed part, once they are held together with a bolt that runs through the center of the part as shown below they are adequately strong to not shear off and make the sides start slipping when the torques of the legs are applied to them.

The star pattern has 6 teeth because this allows for the angle between the two sides to be changed to any increment of 30 degrees. Strandbeest mechanisms that we referenced commonly had the legs offset from each other at 120 degree angles, but our front legs are based on a 180 degree offset because there are only 2 legs in the set. To account for this the part is designed to be able to easily switch angles so we could test different offsets. We ended up finding that a 180 degree offset worked the best because it meant 2 of the 4 back legs would touch the ground at the same time in each leg stroke which helped with stability.