Table of Contents
Prototyping, Engineering Analysis, Simulation
Lab TopicsA preliminary outline has been put together for some of the aspects that could be covered in Dr. Cockburn's lab using this project. The individual modules mentioned below are to be developed separately and basically encapsulate what we will be doing through documentation on this project. This is a first iteration idea and will get more detailed as we get a better idea of how we will go about certain objectives.
Differential Drive ModelThe preliminary differential drive model previous developed was refined to give better results. Following a conversation with Dr. Cockburn, more follow up has to be done to determine if any sort of accurate IMU is possible for the project. We will be contacting Dr. Crassidis in the ME department to make this determination. As of now, a detailed explanation of the current model can be seen below.
After a brief discussion with Dr. Crassidis, it appears that using an accelerometer / IMU will be infeasible for out project due to biasing and error propogation. However, he proposed getting wheel speed data from all four wheels and averaging the speed to get linear speed of the car for the model. We do not yet know if implementing encoders on the front wheels will be feasible, but averaging the speeds of the back two encoders as in input seems quite doable.
Chassis ModificationsProgress on nearly the entire project centers around having a chassis available for testing, so our inital modifications needed to be made to allow for this process. Many areas of the frame were modified in the course of accomplishing this objective, and many will need to be modified iteratively to achieve an optimal design. A succinct summary of these modifications can be seen below.
The spreadsheet used to determine track width can be seen below. This was the design driver for the modifications.
Of course, with these modifications made, a test platform had to be developed, so RC components off of Tim Southerton's car were used to move the vehicle. A video of some movement of the car in preliminary testing can be seen below.
Further modifications had to be made to the chassis after it was discovered that the servo was moving on the mounts relative to the chassis while turning. After some investigation, we decided to rigidly mount the servo more towards the middle of the chassis, which allows for a more robust servo mounting bracket to be used (donated by Tim) and provides easier access to the servo for modifications. Pictures of this can be seen below.
|New Servo Mount Location||Adjusted Optical Switch Mounting Locations|
While the chassis was taken apart, the optical switches on the car were adjusted so that they would have equal gaps of ~3 mm to their respective encoders. This corrected the problem of the right encoder not having a full 5V differential between readings, which was making it difficult to read values using interrupts. Various other minor modifications were also made while this was done so that mounting components was a little easier.
While working on the bumper modifications and since the car had to be taken apart for the weight analysis, the lexan for the chassis was sanded to give a frosted finish. This eliminates the chances of scratching the clear lexan unintentionally, which is a major problem when working with the material. The result looks quite good, and can be seen below.
|Chassis Status 12-1-13|
Course and StorageCurrently we are waiting on the status of available space for Imagine RIT in regards to the course dimensions and parts, but for the time being the track for the Freescale Cup has been identified as a reasonable substitute. The tarp that the RIT CE department has was measured, along with a possible open space in the corner of the MSD meeting area, and these dimensions can be seen below. Half the tarp will fit in the corner with slight wall overlap, and the rest can be rolled out during testing as necessary. We will be working this out with Mark Smith at the start of MSD II.
|Tarp Dimensions||Tarp in MSD Area|
As such, we are currently budgeting 100 ft of tubing in the budget as a possible boarder for the Freescale Cup course, so that users cannot run out of the course area. We will be refining this idea later as more details come in for the Imagine RIT event. The interior of the course can similarly use the tubing as an obstacle. Another possible option is buying plastic cones to signal to the user where the constraints of the course end, but this is not a physical constraint system. Some links to these can be seen below.
The storage of the console was also discussed with Mark Smith, and he agreed that as long as it is on casters it can be stored in bays 5 or 6 of the MSD room for MSD II. Dr. Becker and Dr. Cockburn are still working on where the console will be stored after MSD II.
All of the components for the Freescale Cup Track were moved to the MSD area for testing. These include the tarp, bump, and practice boards. They will be used in MSD II for our testing and for Freescale Cup teams.
ComputerFor the project, it has been identified that using a computer would be the best way to integrate the controls system into the console. As such, getting a separate tower for the console was discussed with Bill Finch in the RIT ME department, and he was able to provide us with a tower at no cost with the following programs installed on the disc image: Google Chrome, Matlab, CodeWarrior, Microsoft Office Suite, X-CTU.
The computer tower will be bolted to the side of the console and the TV will be used as a monitor for programming the interface. A mouse and keyboard will also be provided by Mr. Finch. This way, all the programming and usage will be confined to one computer that is devoted to the project.
The only request made by Dr. Cockburn in this vein is to document what version of the software is used for each part of the programming. This will be done, and I was assured by Mr. Finch that Matlab is backward compatible.
DashboardFor the project, it was identified that signal indicators for the console would be very helpful in making for a realistic driving experience. Freescale has an automotive division that specializes in applications of their boards in cars, so for demonstation purposes they use premade instrument cluster demo boards that look like ones you would see in a vehicle. One of these was seen at the Freescale Symposium, but we were informed that we would not be able to get one of the demo boards. After failing to find any other simply method for getting indicators, we found some other demo models made my Freescale online with documentation, so we followed up and were able to get a MC9S12HY64 for our project. A picture of this can be seen below.
|MC9S12HY64 Demo Board|
Some links to the documentation shipped with the board and those that can be found online can be seen below.
Unfortunately, in order to program the board we need the following items: P & E USB HCS08/HCS12 Multilink Interface Cable, HCS12 Evaluation Board CD. These were not included in the package delivered. We are following up to make sure these are available. However, it does look at though the board will be programmable and of use for our project.
In order to protect the board from users, during transport, and while testing, we decided to use some available MDF as a base and bought some screws, standoffs, and acrylic from Home Depot as a cover. This was the cheapest option that still allowed for easy viewing and access to the side ports of the board. For making this stand, all that was necessary was a mapout of the holes and some dimensions, which can be seen below from Brian's logbook.
|Board Hole Mapping||Mounting Information|
Some pictures of the assembly and final product can be seen below. This is to be mounted to the table surface between the steering controller and TV mount.
A video of the final piece running the preloaded demo can be seen here.
XbeeWith preliminary testing of the Xbee's provided, the following links have been helpful.
The XBee's were tested with the Freescale KL25Z for functionality, and it was shown that these devices are able to be interfaced. Due to changes it the board used it will be necessary to do further testing, but this is an adequate proof of concept.
|Xbee Communication with KL25Z||Arduino Encoder Transmission Setup|
Additional testing was done using the Xbees to transmit data from the optical switches and encoders back to a computer through the serial monitor. An Arduino Uno was used for initial testing and the code used to do so can be found under Software below called "Arduino_Xbee_Speed." Preliminary data taken for wheel speeds can be seen in the Excel sheets below, which use the PLX-DAQ addon to read serial data into the datasheets.
Initial inspections of the data suggest an error in the calculation or sampling algorithm, but the concept is proven without a doubt.
The Xbees was used with some intermediate servo value transmission, which caused the Xbee on the car to become bricked again. After this was fixed, it successfully broadcast both encoder and accelerometer data, which was used for testing.
Range testing of the Xbees was done in the Brinkman hallway in Building 9. A table was setup at the end of the hallway and data was successfully collected from the Xbee all the way to the other end, which was measured to be approximately 141 ft. The Xbees also transmitted through glass to the outside, but the range was greatly reduced. Walls seemed to cause packet loss after the 141 ft mark, but for our purposes this is more than adequate.
Alex and Kevin were able to transmit data wirelessly using the Xbees with the KL25Z in real time using MBED. A video of this can be found here.
EncoderA link to how to use encoders and optical switches for speed measurement in MBED can be found here.
The concept for out encoder design stems from a commercially available speed sensing encoder set that can be found here. This was chosen because it directly measured wheel speed and could be integrated into the design of our 3D printed parts. Some links to alternative designs that Freescale has used in the past can be found here and here. These are not sites with which we are familiar, so nothing was purchased from the vendor. Instead we used the ideas to refine our concept.
The optical gates used for this project were donated and datasheets can be found here. These are reflective optical switches that are to be mounted to the chassis facing the encoder.
The preliminary design for the encoder wheel mount that was printed as a prototype can be seen below. The following conclusions were drawn from fitting the part to the vehicle:
- Torsional rigidity limitations due to delamination of vertically printed shape not an issue here.
- Diameter of inner through hole too tight, but easily opened with #7 drill by hand.
- Hub connection interface very tight, which is good.
- Encoder rigidity good, but due to changes in optical switch it should be moved toward the hub and made larger.
- Hex nut on the end accurate but could be made larger for a more snug fit on wheel.
Using the above notes, a new encoder mount was designed, which can be seen below.
Some images of the mounted encoder and optical gate can be seen below.
|3D Printed Encoder Prototype||First Cut Mounted Optical Switch|
The second design of encoders were 3D printed and fit to the car with minor modification. The optical switches were then mounted to the chassis. An image of these changes can be seen below.
|Second Design Encoder and Optical Switches|
A video of the performance testing done can be found here. Encoder data was taken after the optical switches were adjusted. The car was set at a dead stop and accelerated at full speed before braking and reversing at full speed, coming back to a stop at the starting point. The data collected yielded good results that were as expected, but there were a number of outlier points caused by noise in the collection method. We believe this can be corrected using a derivative smoothing method.
|Filtered Forward Data||Filtered Reverse Data|
|Wheel Speed Data|
From the above data, we were able to gather the following statistics about the vehicle movement. These compare very favorably with those outlined in our metrics.
|0 to max speed||2.8||s|
|max speed to 0||2.4||s|
Console SteeringA Logitech MOMO force feedback USB steering wheel and pedal assembly was donated to the project by Tim Southerton, so anaylsis was done to verify how it could be used. Initial testing showed that the Logitech driver, which is outdated, buggy, and only available for Windows, was not an optimal design path. Writing a new driver was deemed to be a difficult prospect, so we looked into putting analog data out of the control components and into a microcontroller like the KL25Z. Some diagrams of the internal wiring were located, which can be seen in the schematics below, which showed that the pedals were simply potentiometers that supply analog signals. The steering wheel console was opened to determine operating mechanisms. From the images below, one can see that a rotary encoder and a optical switch work together to give displacement data of the wheel.
|Console Internals||Steering Optical Switch|
|Steering Encoder Mechanism|
Diodes that given directional data were located on the board. Oscilloscope data was gathered on this interface and the one for the encoder, which can be seen below. This determines proof of concept of reading the values into a microcontroller.
|Encoder Scope Data||Optical Switch Scope Data|
Additional notes on the testing can be seen below and in the Photo Gallery.
|Notes Page 1||Notes Page 2|
From the analysis, the following conclusions were drawn.
To allow the EE's easy access to the electronics on the steering controller, an access hatch was cut into the front of the console. It was decided that using hinges and a magnetic cabinet door mechanism would make for a good access mechanism, so these items were added to allow the plastic to be easily opened and closed for testing. The KL25Z board was also mounted in the enclosure by removing some of the fins from an air vent to allow easy wiring of the board to the console electronics. All parts for this modification were donated by Tim. Some pictures can be seen below.
|Console with Access Hatch||KL25Z Board Mount|
|Hinge Mechanism||Magnetic Cabinet Door Mechanism|
Matt and Lalit were able to tap signals out of the console and into the KL25Z. The USB connection was also hot glued for protection of the wire which tends to move around otherwise.
The general approach for the data flow at the console was complicated by the inclusion of a computer, so flowcharts were done of the car and console systems to show what is going on. These can be seen below.
|Car Flowchart||Console Flowchart|
Timing GateA timing gate is necessary for our project to provide a competitive element to the racing events and allow for measurement of the car performance to meet the specs and compare with the encoder data. A preliminary frame was built for Proof of Concept testing, which can be seen below.
From this setup, analysis was done to verify that it can be used as a timing gate for our project.
Bumper DesignTo support preliminary testing, ideas for car bumpers were established so that the vehicle could be driven without risk of damaging the components. As of now the best temporary option established is adhesive foam mounted on a rigid bumper, which is currently made from PVC, metal standoffs, L-brackets, and bolts. After some preliminary testing, it can easily be seen that the design works but is not robust, as the bumper bent after a low speed collision with a couch.
|Testing Bumper||Bumper Deformation after First Crash|
A more professional long-term option that has been identified is a machined nylon bumper designed to absorb and dissipate impacts. This is of minor risk so will be addressed in detail later.
After the bumper was broken a few more times, the metal for the L-brackets was changed to steel from aluminum so that it would deform less. This works fine for testing, but it has been identified that the final bumper needs to be wider so that the front wheels cannot catch on other objects as it is driven around. The bumper also needs to be at a higher level so that the car cannot drive the bumper under objects and still hit the plexiglass off of things. The current bumper is also too low so that it catches when the car is driven over the bump. These testing items can help design an optimal bumper later on.
Below is a link to a foam that might be useful for the final bumper. Brian found this and will be working on characterizing the bumper material.
Below is a document Brian used to characterize the foam needed for making a bumper. It effectively uses an energy approach to a worst-case-scenario wall collision to estimate density needed.
Based on these calculations we have decided to purchase some polyethlyene foam and threaded rod from McMaster for fabricating this bumper. A preliminary bumper design can be seen below. This will be further modified after the foam received is characterized. The parts for this come to around $10. The lexan for this has been donated by Tim as there was a piece available and it is only $3.48 at Home Depot.
|Second Iteration Bumper Design|
In order to do anything with this foam bumper design, a hot wire foam cutter had to be made, as this is the only way to create nice, custome cuts on foam. Because this is an extra expense, Tim has always wanted one, and he bought the parts and is keeping the tool, the item was build using online directions. A good tutorial of this can be seen here , but that is all the detail that will be given for this non-project item.
With the foam cutter built, which can be seen below, cardboard templates were made for the bumper. Being as the foam purchasing was turning out to be quite expensive, we got lucky finding some free polyethylene foam pieces in bay 3 of MSD, so these were used. The cardboard pieces are pinned to sandwich the foam, and the cuts are made when the wire is hot. A picture of this can be seen below.
|Hot Wire Foam Cutter||Cut Foam and Cardboard Templates|
The bumper was made by machining the lexan using the schematic below for the bottom support. For the top support, one inch of overlap was left for mounting to the chassis standoffs for support. Threaded rod was chosen as the only valid bumper supports, so these were cut to length and flats were machined in for tightening using a Dremel tool. The foam was inserted into and secured between the lexan frames, and these frames were then joined to the chassis as seen below. The bumper mounting holes had to be bored to clearance for the threaded rod.
|Threaded Rod||Foam Sandwich||Chassis Bore|
After some preliminary testing, it was determined that the foam was too stiff for the car, so reliefs were cut into the foam for stiffness reduction. This worked rather well, and the new bumper shapes looks notably interesting.
|New Bumper Design|
Console ChairSignificant thought and effort has been put into the console chair, as this item will really make the project memorable. To provide an authentic driving experience, a chair from a real car was identified as the only accept option given the project name. Getting a seat from a car became a target for proof of concept, as the only affordable options are used seats. FMS was contacted to see if they could help out on this item, and Chris Furnare and Jim Shuffield were happy to provide us with a chair from a '02-'03 Chevy Venture. It was decided that a turning platform would make getting into and out of the chair easier, so an old office chair base was donated to the project by Tim Southerton. To assist with design, a CAD model of the assembly was made. This can be seen below.
The desk for the project was also obtained for free because it was being thrown out, so most of the components were available for the construction. To reduce cost, a simple 2"x4" wood design was established for a preliminary test bench, and the wood was provided by MSD (free from Bay 3 in the Design Center). Similarly, free bolts, nuts, washers, and screws were found in storage cabinet 1 in the design center. The MDF for the supports was provided as scrap material from Tim Southerton's Furniture Elective class, and were cut on building 7's saws (thanks Will).
To clear up all the items that were accumulating in our apartment, Brian and I went about assembling the preliminary design for the chair, which took all of about 7 hours. A picture of the first iteration can be seen below. Modifications need to be made for easy adjustment, a more professional appearance, and easier mobility, but the concept is clearly proven.
|Console Chair 10 27 13||Console Chair 11 5 13|
Doing some follow-up with comments from Dave on the Baja Team from out last review, we looked into pneumatic casters from Caster City. The smallest we found were 6" casters with a ride height of ~7.5", which is significantly higher than previously estimated. However, such large casters would allow for easy transport, and the Baja Team might be able to get them for free as Caster City is one of their sponsors. 6 of these casters are valued at around $150.
With this possible change in height of the console, some modifications had to be done with the original chair. The height differential between the seat and the table base was still an issue, so we decided to move the slider mechanism to the table mount so that the table could be slid toward the operator by the team member helping with the demo. The chair base that then bolted directly to the 2"x4" rails so that it was more stable. This allowed the chair and table adjustments to move over an acceptable range for most operators.
Metal Z-brackets, probably 1/4" in thickness, were identified as a mounting option for the casters. We are still waiting to hear back on what the availability of the casters will be before we go about purchasing or designing this functionality. The current state of the seat can be seen above.
Following some additional conversation with the Baja Team, some cheaper casters were selected so that the team has a backup set to ask for if the pneumatic ones were too expensive. These casters can be seen below.
The casters are going to be in a batch order with other Baja Team projects. They can be expected around the start of MSD II.
A seatbelt was also locates for the chair, which may be added to the final console for a more realistic feel.
Further modification was done to the console, as 2x4's were cut at 34" long and lag bolted to the frame in three places as caster mounts. Casters taken off of the original table were temporarily mounted to these cross bars so that the desk can be moved around. The lag bolts were free from Tim.
The table was also deconstructed, and after talking with Rob Kraynik from the machine shop, it was established that better table surfaces were available for the desk. We are currently working on getting these cut to size and installed, but the result should give a more professional appearance to the project. However, more supports will be necessary to keep the desk from oscillating.
Spare angle iron from the MSD area and Tim's supplies was used to create a TV mounting bracket that can be attached to the table surface. This allows for additional open area on the desk surface, as the angle iron mounts on the underside of the table.
It is estimates that all of the components can be assembled at some point during the first few weeks of MSD II. Once the computer is available, this will also be rigidly mounted to the bottom table surface. Pictures of the modifications made can be seen below.
|Caster Crossbar Mounts||TV Mounting Bracket|
The pieces of the console were put together following the location of a piece of plywood as a connecting piece. The resulting desk looks great and is quite rigid due to the amount of bolts that were used in the construction. The results can be seen below. Possible areas for additional work include:
- Edging for the table surfaces
- Bolting the computer tower to the front caster crossbar
- Connecting the crossbars on the outside edge of the base to prevent flexing and bowing
- Vibration dampening of the table
|Updated Console 12-11-13|
Camera TestingAfter much discussion, it was decided that an analog output camera transmitted over analog RX/TX set was the best option for wireless video. The camera tested to verify proof of concept was donated to the project by Matt Morris, a mini helmet camera with little documentation. The video TX/RX was donated by the RIT ME Robotics Lab, and contains little documentation as well. Some work needed to be done to make up connecting cords and power supplies, but acceptable voltages were determined and signals were transmitted. With some adjustment the image is quite good for free components.
Following a discussion with Dr. Becker and Dr. Cockburn at the preliminary DDR, a weight analysis was done on the car to verify that the final weight of the vehicle with all the added components would not be too much for the motors to handle. This can be seen below, but the results suggest that the car will be within 10% of the total weight of the Freescale Cup car from last year and 20% under the weight that the car was when we tested its performance, so weight will not be an issue.
Another issue posed for the DDR was if all the additional components would fit on the car, which involved some additional CAD work. The results can be seen below, which suggest that everything should fit with ample space on the adapter plate so that future revisions could make the package even smaller.
|Car Component CAD Mockup|
Similarly, the power budget was modified to include the new information from the camera components. This can be seen below, but the overall conclusion is that the batteries will be able to power the car and camera in excess of the desired 1 hr operation time between battery changes.
CAD, Schematics, and SoftwareThe following directory contains the CAD drawings created thus far for the project. Thanks should be given to GrabCAD.com for parts in the creation of the console assembly and car components.
A detailed electrical plan for the project can be seen below. This document summarizes and documents the approach that will be taken with modifying the TFC motor shield and making a perfboard shield for the Xbee and encoder circuits.
The software below includes Arduino code made for preliminary testing of POC, along with some libraries from MBED designed for the Freescale Cup Car. These documents can be downloaded here.