P16029: Robofish 3.2 - Object Retrieval

/public/

Preliminary Detailed Design

Table of Contents

1 Team Vision for Preliminary Detailed Design Phase
2 System Flow Chart
3 Image Processing
4 Hydraulic System
5 Buoyancy Control
6 Jaw Design
7 Return Home
8 Bill of Material (BOM)
9 Fish CAD Model Design
10 Risk Assessment
11 Plans for next phase

Team Vision for Preliminary Detailed Design Phase

The objective of this phase to was finalize many open questions that remained for the fish.

Questions such as, "How many pumps?", or, "How will the fish catch the object?"

What was significant for this phase was the selection of the pump.

With the new pump purchased and delivered to our team, we were able to successfully run the pump and gather data.

With these conclusions, we were able to select how many pumps would be needed, which we have determined as only 1 will be required to complete all the Customer Requirements for both teams.

Note, these were not the only questions asked, or answered.

The Design Review should mimic the flow of how the Robofish will complete the task.

Tasks for this Review

Timeline and Completeness

System Flow Chart

Overall System Flow Chart

The Flow Chart above details the process the Robofish will make to intercept the object.

There are a few key assumptions made in this chart for simplicity.

If the fish does not catch the object, it will assume that it has hit the bottom of the environment and will return back to home.

The dive stick will be standing straight on the bottom of the pool, not laying on the bottom of the pool.

Image Processing

In this phase we started looking into the different options for cameras that could be implemented for the image processing, the two primary options were either a regular webcam, or one with a fisheye lens. In order to get a better idea of which camera is better for our system, the field of view of both were calculated. Further discussion is needed, but in terms of cost effectiveness and ease of implementation, the regular webcam appears to be the better option.

FOV Estimates

Utilizing the input from last review, the algorithm for detecting the object was modified to now identify it using color versus "teaching" the program what the diving stick looks like. The use of tracking the color allows for greater versatility of the program, because now it can track any object as long as it has been set for the appropriate color.

Object Detection Algorithm

To determine the distance to the of the object, it was chosen to use the relative size of the known object at known distances to create an equation that when inputted the pixel width will give the relative distance of the object. This was tested with measuring the size of the object every two inches until the object was three feet away, and the given data for the current setup resulted in the equation in the graph below. When the equation was put into the program it outputted distances within two inches of the correct distance.

Distance Using Width in Pixel

Video of the program outputting distance can be seen here, and of the Raspberry Pi communicating with an Arduino can be found here

Hydraulic System

We have changed our hydraulic system to a simpler design. Instead of using a compressed air tank to force water back out of our ballast tank, we will use a series of valves that can redirect the flow from the pump. This lets us draw water from the environment to fill the ballast tank, or draw water from the ballast tank and push it back into the environment.

System Hydraulic Schematic

We ordered a pump that provides 5 L/min of flow and a max pressure of 100 psi. We were unable to create an experimental pump performance chart due to the difficulties of measuring the pressure of a diaphragm pump. However, we did measure the flow rate through one of our valves to be 2.6 L/min, which corresponds to 48 psi of pressure on the theoretical pump performance graph.

Pump Performance

Using this initial pressure and the slope of the pump performance graph, we modelled how the ballast tank would fill over time as the air inside it was compressed. From there, we estimated velocity, depth, and the time to intercept a sinking object.

Intercept Calculations

By changing the size of the ballast in the calculations above, a chart of the time needed to intercept a sinking object at two different speeds was generated. Based on this chart, we were able to determine an appropriate size for the ballast tank.

Ballast Tank Size

Buoyancy Control

Fill and Drain functions

Buoyancy Control Flow Chart

Method1: Hydraulic Mock up

Method1: Current Pulled by Mock up

Method2: IN-3/8", OUT-3/16"

Method2: Current Pulled by IN-3/8", OUT-3/16"

Method3: IN-3/8", OUT-3/8"

Method3: Current Pulled by IN-3/8", OUT-3/8"

Each Methods'Performance

Jaw Design

gears not included on CAD design

Jaw Design - Front

Jaw Design - Back

Jaw Design - Side/Bottom

Why did we choose this design?

Easily attachable to the body of the fish
Easy grabbing of objects
Light weight, with intentions of 3D printing the parts
Low drag force
Able to pinch objects of multiple shapes
Easy attachment of McKibbon muscles on back of jaw

Return Home

RSSI was monitored for 450ms then histogram was generated.

Variation of RSSI at a fixed distance of 3m

Histogram of RSSI variation

Simulation of approaching "Home" was conducted by moving cell phone towards the beacon in constant speed. It indicates that "Home" location can be approached by using RSSI of a single Bluetooth beacon within 5 meter radius.