Final Demo Video

Its been a great semester! Here's what we built.

The Reach Goal: Multiple gaits and closed loop

For our reach goal, we decided to implement multiple gaits for the Soft Stepper (wrapped in an API) and implements a simple closed loop in the system.

In addition to the our basic walk, we were able to implement a faster (and far more complex) "power walk" gait, as seen here.

Additionally, we were able to implemented a turn gait.



Currently, our closed loop system is a collision avoidance module. This required the use of a ping sensor and wifi module connected to an additional mbed which was installed directly on the Soft Stepper.

Soft Stepper Demo Video!

It walks!

Soft Stepper: the Baseline

In order to build the Soft Stepper, we were required to design another generation of actuators. Imbued in this version is the sum of our experiences (read: failures) of previous generations. And this time, we got it right!

This generation is short and has relatively shallow cavities like generations 2 and 3, but has larger, divided cells like those of generation 1. The result is a functional mix of stability and flexibility. 

Additionally, in order to allow us to continue using the chassis, the newly created mould has a cross section identical to those of generations 2 and 3. 

The final Soft Stepper walks slowly but surely. Demo video to follow!

Soft Stepper Update 1: The Control Board

Over the past few days, we've made significant progress on some of the most crucial components of the Soft Stepper.

New moulds were designed, printed, and set to cast. We initially designed a shorter, 6 chambered version.

Unfortunately, a mixed axial/bending stress analysis revealed that our application required actuators that were longer and contained more chambers.

Thus, we created a new set of moulds and set the legs to cast in these. 

As soon as they are done and cured together, we will attempt to incorporated them into the chassis we designed and printed on the Makerbot.

Additionally, we designed and laser-cut a control board for the components.

The pump and solenoids were immediately integrated onto it. When the PCB is received, we will be able to add the mbed, pressure sensors, and mosfets. 

We created a short demo video to showcase our recent developments, implementing the rhythm for one of the Soft Stepper gaits. 

Final Project: the Soft Stepper

Since our soft fish idea went down the drain (no pun intended) we will move on to a new idea which will build on our previous work with soft actuators in a more fundamental fashion.

Our proposed project is a take on Harvard's PneuNets Robot for multiple modes of locomotion.

Our robot, the Soft Stepper, will be an adaptive multiple gate walker capable of moving along complex paths using soft-actuated legs attached to a rigid body chassis. 

The legs will be cast from a mould design similar to the fish tail, with enclosed cavities.

The walker will be controlled by an mbed mounted on-board via PCB, which will also contain the pressure sensors. The pump and solenoids, on the other hand, are much heavier and will likely be hosted on off board. 

In the final presentation, we will hopefully be able to demonstrate the versatility of this system by moving it in different directions, running it over terrain, and possibly altering planes of movement (squatting) or different speed settings.

The Fish Fail

Our soft fish cured beautifully and the fillets bonded well to the interface.

Unfortunately, the actuators were far too rigid, and would not bend far enough nor with enough force to simulate swimming. 

Baby Fish: Update

We've designed and printed molds for the actuating section of a pneumatic fish, pictured below. These are: the outer shell mould (white, center), the cavity mould (purple, left), and the restrictive interface mould (Purple, right).

in addition to the cavity mould shown above, we printed another, mirror image mold to compliment the existing half and for a full tail. The pneumatic actuators cast from these moulds, or "fish fillets" as we call them, are both done, and currently curing to the interface. Hopefully, we will be able to observe their functionality later tonight.

Today, we received some components crucial to the progress of our project. This includes such items as 1/16" ID tubing, and m5 sub-plates and connectors for our valves (pictured below). 

Over the next couple days, we intend to determine a configuration of electronics that fits inside the fish form factor and provides adequate functionality. We will do this by experimenting with the current tail, interfacing through a connector plate (yellow, top in first picture) until a more permanent container can be designed.

Planning our Reach Goal: the Soft Robotic Fish

Following today's discussion with Professor Mangharam, we will be pursing the creation of a soft-actuated fish that would be able to swim underwater. This project is based loosely on this MIT research project.

Essentially, this fish would have three primary segments: The head, the midsection, and the rear.

The rear of the fish is comprised of two large specialized actuators, bonded together. In between them lies a rigid piece of plastic or hardened silicon, which can be further extended and shaped to form a tail. 

We will conduct our first mini-experiment in this area by fabricating a small version of this fish, with a body diameter of about 2 inches and length of 4 inches. 

The pneumatic channel will be part of the mold and run the length of the rear section adjacent to the middle substrate. 

The midsection will interface with the rear via a bonded silicon seal plate paired with a matching acrylic faceplate. The silicon plate will include extended walls to envelop and thus water-seal the acrylic faceplate. 

The mid section and part of the head will include our electronics, likely including an mbed, pressure sensors, solenoid valves, and possibly and exhaust port/swim bladder.  A potential arrangement of these parts is displayed below. 

With a small amount of room left in the head section, it should be possible down the line to embed some form of sensor or other item which we might find useful or interesting. 

Baby Step 1: Basic Pneumatic Gripper

For our first mini-project, we connected two pneumatic actuators to a pump, with valve for release, and created a soft gripper. The final configuration is diagramed below.

We initially planned to utilize all three of our actuators, but one (the original, defective cast) popped as we attempted to actuate it. Fortunately, two of these were sufficient to complete simple tasks such as holding a cup (see video). 

Synthesizing our First Pneumatic Actuator

We started off with Dragonskin 30 elastomer. This material comes in two separate liquid components which begins to solidify when mixed (1:1). 

This mixture was then poured into two separate casts: The cavity cast (above) and the cell cast (below), both which we printed on MakerBot Replicator 2s using these .stl files at this link.  

When casting the base (top picture), it was necessary to place a piece of paper within the silicon. This was done to ensure the rigidness of the based upon actuation. 

When we tried casting our first actuator, a lot of silicone leaked from the crack between the mold halves -- the website says nothing about this. After emptying it out to try again, we noticed that the silicone that was left near the edges cured in the meantime, preventing new leaks.

A video of the casting process: 

After pouring the silicon, it must be left to cure. Although the material specifies 16 hours at room temperature, our application takes closer to 6. 

After drying, the actuators were coerced out of their molds, glued together with additional silicon, and tested... and it worked!

Basic Modes of Touch- a set of experience goals

We are approaching our project by build atop existing soft robotics architectures defined by the Harvard Soft Robotics Toolkit. In selecting this project, our goal is to use this platform to build a basic set of universal experiences which can be applied as modules in future soft robots. In order to do this, we require a well-defined set of subsenses comprising the primary sense of touch, also known as the somatosensory system. According to Wikipedia,

"Although touch (also called tactile perception) is considered one of the five traditional senses, the impression of touch is formed from several modalities including pressure, skin stretch, vibration and temperature."

These criteria of pressure, stretch, vibration, and temperature can serve as the initial bases of our experience set.

Elastomers and Pneumatics - Shopping

We found out that the elastomers can cure at lower temperatures (but slower), so if the mbot trial works well we can use those for actual molds from here on out. The blood pressure cuff has some complicated details involved, so we're going for a three-finger grasper first, and we'll use the blog to detail that.

A couple immediate additions to the material list are the elastomers: Ecoflex, Elastosil M4601, Sylgard 184 (aka PDMS Silicone). Those sites are the "official" sources, but if there are any cheaper sources then we should use those. We also need access to an oven for curing (40-45 C is the goal) and a vacuum chamber (which could arguably be constructed from a shopvac and a cardboard box).

These pumps are 3V, 200mA so we can probably drive them directly via H-bridge, for testing the actuators under pressure. To actually control them we need the three-way solenoids to allow for exhaust, which implies a valve manifold for organization, which implies a stronger monolithic pump. The documentation calls for this pump, but I've found some from a previous incarnation of the same manufacturer with comparable specs, at about 10% price. The listed solenoids are actually quite good, the next best option require a high minimum pressure differential to function. That manufacturer has nice manifolds, connectors, and tubing.

- 3-way solenoids = $35 each (same as official)

- Solenoid manifold = $45 (same as official)

- Pressure sensors = $20 each (vs $45)

- Air pump = $35 (vs $300)
- NPT to barb coupling = $1 each (no minimum order)
- Barb 'T' coupling = $1 each (no minimum order)
- Tubing = $12 for 30m

Brainstorming

Toolkit source: http://softroboticstoolkit.com/

- Master pump, control valves/release to components

Detection
- Embed sensors in material while casting

Motion Experiences
- Fiber-Reinforced actuators: contraction, flow experience
- SDM fingers: contraction, squeezing
- PneuNet Bending actuators: "stars" can simulate local squeezing/experience
- Dialectric Elastomers: temperature change via insulation between a heat source
- TakkTile Sensors: small pressure sensor, embed in soft components, e.g. assist in guidance

Development Priorities
- Ari flow control (pressure sensors, pump actuation, airflow couplings)
- Developing modes of feedback (active, passive)
- Sensing, incorporating into system