Iteration 1

In the initial design phases, we decided to proceed with a leg structure analogous to most mammals with a femur, a tibia, and a foot (Fig 1 , Fig 2). 

Open

Fig 1

In order to simulate a bio-informed tendon compliance, we first had strings attached to a motor as an actuator that pulls on the links, and linear springs that recover position. However, we weren’t confident in analyzing a four-bar mechanism with transient link lengths due to the string. We also couldn't justify this design that requires a motor that delivers the amount of energy at the scale of the package we are constrained with, due to weight considerations. Additionally, we have the foot composed of a tarsus link and a toe link, which means an extra DOF and unnecessary complexity.

Iteration 2 (Delivered Initial Prototype)

For our next design, we decided to use torsional springs as actuators for their energy density compared to motors, and simplicity of output profile and integration to rotary outputs compared to tension/compression springs. Compared to our previous design, we also doubled the tibia to make a four-bar with only one DOF (if grounding link1). The curved contact foot design was inspired by the sprinter's prosthetic. We believe it adds overall compliance and more consistent contact with the ground as the entire leg rotates and pushes forward.

Here’s the first manufactured prototype we made (Fig. 5). The device is capable of leaping forward with the help of a human hand on compressing the spring, which proves our concept’s viability. It also has the potential to fit more and stronger springs to expand its actuation capabilities. For our next step, we need to consider methods of storing the required elastic energy in the spring. 

Open

Fig. 5 Initial Prototype

Iteration 3

In this iteration, we flushed out many details and locked in with the overall design to proceed with final manufacturing and optimization.

First, after consultation with the TA, we decided to remove the compliance with a stiff link by thickening the foot's effective bending area: our application doesn’t involve protecting knee joints and recycling the energy upon the foot’s impact into the ground when landing. Thus, we decided it’s easier to keep the foot link as a stiff link. 

Secondly, to address the auxiliary energy source issue that stores elastic energy into the torsional springs, we came up with a cam and follower mechanism design that utilizes a continuous servo motor to realize spring compression and release intermittent motion (Fig. 5). We opted for two servos that power each leg separately in complete mirror to avoid the need to design around a transmission system and worrying about one motor not capable of powering two cams. (However, the syncing challenge between the two servos became a major issue in our final demonstration)

Thirdly, considering our overall system is a rather complex dynamical system, we want to reserve the option of fine-tuning its behavior after assembly to play with its balance. Some of the tunable parameters incorporated into the design include: (Fig 6)

1. Overall weight distribution adjustable from battery placement.

2. L1 angle (Leg module’s relative angle to chassis, when chassis is global ground in 4-bar analysis). This is crucial in distributing thrust in vertical and horizontal direction, which directly affects the system’s hopping behavior. 

3. Cam size. Determines spring compression capacity and complements changes in L1.

4. L2 extension length. This complement changes in L1. 

5. Less of a tunable factor, but the leg springs are relatively easy to replace to experiment with different effective overall spring constants. O4 is also designed to fit springs in case we need to add more actuation capacity.

6. Foot. Swappable with different foot designs and configurations. For example, we ended up using the “duck foot” configuration to ensure maximum lateral stability.

Below is the completed final CAD model

Open

Isometric Views

Draft Bill of Materials

LC* : laser cut

3P* : 3D print

Reference

Hand Sketches