Last updated: 5/9/2018
This is a summary of most of the mechanical/mechatronics work I led on the MARTY research test platform from 2012 to 2018. The information and pictures below show the state of the project circa May 2018, and are organized by sub-assembly.
This page is a work in progress. Cognizant of just how much I've learned from the stories and pictures of others' projects on the internet, I've decided to really try and put this stuff up. I really wish I had started a blog in 2012 -- this is my attempt to historicize after the fact!
Quick Links: Summary | Timeline | Drivetrain/Rear Battery Pack/Power Electronics | Front Suspension | Steering/steer-by-wire | Front Battery Pack | Interior | Rollcage/Seats/Harness | Rear Suspension | Brakes/brake-by-wire | Measurement/Compute
MARTY is a heavily modified 1981 DMC DeLorean, that is used by my colleagues and I at the Stanford University Dynamic Design Lab (DDL) to study vehicle dynamics at and beyond the limits of stability. Specifically, my PhD research focuses on autonomous drifting.
I have had the amazing opportunity to lead the development of the vehicle since the project's inception in June 2013. As of May 2018, the major modifications to the car include:
- 700V/300kW/20kWh Renovo.auto electric drivetrain, with rear and front lithium battery packs
- Coilover conversion + custom front and rear suspension and uprights
- Full rollcage, bucket seats, and six-point harnesses
- Steer-by-wire + upgraded steering rack and steering range
- Big brakes + individual corner brake-by-wire
- Revamped interior
- Integrated dual-antenna RTK GPS-IMU unit for motion measurement. For autonomous calculations, primary hard realtime automotive control computer, plus secondary Linux-based compute network.
|Jun 2013||Car purchased, benchmarking begins.|
|Aug 2013||First implementation of steer-by-wire + GPS-IMU/computer integration completed.|
|Dec 2013||A simple throttle-by-wire for the gas motor was completed. First Thunderhill test.|
|Jan - Apr 2014||Removal of stock drivetrain, and separation of body and frame.|
|Apr - May 2014||Design iterations for electrification phase 1 (Renovo drivetrain + 350V rear battery pack).|
|Jun - Sep 2014||Execute hardware plan for electrification phase 1: frame modifications, mounts + motor cradle + rear subframe manufacturing, and test fitting. Design + build for rollcage/seats/harness.|
|Oct - Dec 2014||Integration of electrification phase 1: wiring + electronics, installation, testing, testing, testing.|
|Feb 2015||First Thunderhill trip for electrified MARTY! Begin coilover conversion for front and rear suspension.|
|Apr 2015||First Thunderhill test with new suspension. First basic autonomous shakedown tests.|
|Jun 2015||First autonomous drifts (no path-tracking).|
|Aug 2015||First test with new steering rack. First autonomous drifts with path-tracking!|
|Sep 2015||First Thunderhill tests with revamped interior. Filming for unveil video.|
|Oct 2015||MARTY is unveiled at an Open Garage Talk hosted by Jamie Hyneman.|
|Jan 2016||First MARTY conference paper submitted (IEEE Intelligent Vehicles 2016).|
|Mar 2016||First cassette-tape drift test!|
|Jun - Sep 2016||Design and build for front battery pack. Design for new front uprights + brake-by-wire integration.|
|Oct 2016||"Cloverleaf" at Golden Gate Fields. Begin build + integration for front uprights + brake-by-wire immediately after.|
|Nov 2016||First test with front uprights + brake-by-wire.|
|Mar 2017||First successful tests with nonlinear model inversion + wheelspeed control.|
|Jul 2017||First successful tests with online parameter estimation.|
|Aug 2017||First really solid figure 8's with nonlinear path planner!|
|Apr 2018||"Dr. Octopus" test - what is Dr.Octopus? It shall be revealed later...|
Drivetrain/Rear Battery Pack/Power Electronics
The electric drivetrain, consisting of the motors, transmissions, front and rear battery packs, power electronics, and management + remote telemetry computers/software, were designed and built by Renovo.auto. The Stanford team, with guidance from Renovo, designed and fabricated the packaging solution. The initial 350V/150kW/10kWh system was installed in 2014-2015, and the front battery pack was added in 2016 to bring the system up to 700V/300kW/20kWh.
Designing the packaging solution for all of this stuff really began with taking the car apart in Spring 2014. The DeLorean is a body-on-frame vehicle, so we could, after removing all the plumbing, wiring, and bolts that connect the two, pop the body off on the lift. This allowed us access to measure and CAD the frame, and then iterate packaging solutions on the computer.
We wanted to move quickly, so rather than trying for precision measurement and modelling of the frame everywhere, we accepted rougher tolerances in most places; our packaging design then drove which measurements needed to be more precise. Paper prototyping - or really, laser-cut-duron prototyping - then allowed us to verify our fit and clearances. I believe that this iterative, prototype-based process really helped speed things along, as we were learning a lot with each cycle - this is especially true for us students who had never done much work on cars before, not to mention a drivetrain swap!
The motors and single straight-cut gear transmissions are mounted back-to-back in the 'motor cradle'. The aluminum plates of the motor cradle were designed by Stanford, and verified + manufactured by Renovo. We used subframe bushings originally from the BMW E39, pressed into simple holders that were then bolted to the cradle.
After quick prototyping with L-channel mock-ups to determine final dimensions, I designed and machined the cradle mounts. This was tons of fun as we had a Haas VF-1 in the shop, so one could quickly make some holes in a prototype and test it out there and then. It also had a tool length probe and workpiece probe, which I was using for the first time - and it made me giddy with power. Structural FEA helped drive the final design, though we certainly did not weight optimize very strongly, electing instead to err on the side of higher safety factor and faster overall progress.
The rear battery pack is mounted inside a tubular steel subframe. This is in turn bolted to the tapped bungs welded into the rear cross member. The power electronics modules (PEMS) are attached on top of the rear subframe. The motors, power electronics, and charger are all water cooled. This was a pain-in-the-ass - as Calvin's dad would say, 'character-building' - but also really educational, especially with expert guidance and tutelage from Matt Brunner at Renovo.
This electrification project was a big one. From the Stanford side, assisting with all of this were the legendary OG MARTY Master's students, Phill Giliver, Shannon McClintock, and Arni Letho, and Wyles Vance who helped with the welding. This project was a huge learning experience for all of us, and we certainly have to our crazy-awesome partners at Renovo.auto not just for their wonderful drivetrain, but for sharing their suggestions, design reviews, and experience, in particular: Matt Brunner, Aaron Sellars, Connor Sullivan, Mike Vogel, Nick Hori, Jason Stinson and Chris Heiser. Thank you for all the late nights we spent together cutting, welding, CADing, machining, bleeding brakes, installing and reinstalling motors/battery packs, bleeding coolant, and drinking beer!
Custom A-arms + coilover conversion
One of the first things we discovered back in January 2015 after first installing the drivetrain and hooning MARTY v0.5 a bit, is that the stock DeLorean setup is really not very good for drifting or racing - or really performance or fun of any kind. Namely, the static camber + camber loss was profoundly off, and we were using about an inch of the outside front tire on turns, leading to truly epic understeer all the time.
After analyzing the existing suspension geometry, we were pleasantly surprised to find that lowering the ride height would achieve huge improvements in dynamic camber, and that adding a little negative static camber on top of that would get us to where we wanted to be. (DeLoreans riding too high is actually a known common problem. ) Making our own adjustable-length lower A-arms + strut tower modifications to package coilovers was a good way to accomplish and tune all of this while also 1) adding easy spring rate/damping adjustability and 2) improving our ability to take the higher structural loads from agressive maneuvers.
Most of the suspension analysis was led by Tushar Goel, who was then a first year Masters student, and had done a ton of this stuff with Formula SAE in undergrad. There was an element of 'My First Custom Suspension' to the whole mechanical design and fabrication process - thankfully, Aaron Sellars of Pacific Custom Fabrication/Renovo.auto was there to guide us. From finding local ball joint vendors, to design reviews with years of automotive experience behind them, and even to doing the actual welding itself, Aaron really helped us out and taught us a ton.
Custom uprights + hub switch
After destroying a rough average of 1 wheel bearing every two months for a year and a half - this was even before the frunk pack, so the weight on the front axle was the same as stock - we finally decided to overhaul the upright assembly as part of the big frunk pack + Brembo brake-by-wire upgrade.
To that end, in Fall of 2016 we designed an upright to use a S550 Mustang front hub. The modular design allows great adjustability, especially for the steering arm. This was great fun to design, as by this time we had done a great deal on the car and were getting pretty confident. 3d printed prototypes helped us verify all the clearances. In the interest of saving time, the final product was partly machined by us, and partly by an outside shop, for the first time in the MARTY project.
We have not broken any more front wheel bearings.
Pictures of current system coming soon!
Steer-by-wire was the first system we developed for MARTY, back in 2013. The original DeLorean did not have power steering of any kind, but the strong enthusiast community had come up with a solution: Ed Uding of Delorean.eu makes a bolt-in steering column with electric power assist. This caught our interest, as from the pictures of the system, it looked like it would be straightforward to replace the stock motor with a more powerful one that we could interface with. And so this is exactly what we did! We machined a custom adaptor to bolt on our much bigger 48V brushless motor to the column, and got a custom spline coupler cut with wire EDM.
We use a CANOpen absolute encoder on the back of the motor, which makes control and calibration an absolute breeze. Motor control is also achieved with a unit that follows the CANOpen spec. Pre-electrification, power was simply provided through a separate 48V lead-acid pack. Post-electrification, we now use a 12V-48V DC-DC step-up, from our 12V circuit + lead acid battery, which is in turn charged by a 2.2kW 700V-12V DCDC convertor from the main lithium pack whenever the car is energized. Low-level steering angle tracking is achieved using a lead-lag controller with a feedforward current computed from the desired front lateral force. When initially sizing out the system, I somewhat surprisingly had the foresight to greatly oversize everything. This was fortunate: we now have double the stock weight on the front axle, a faster steering ratio, and when we do Figure 8 transitions we want to go lock-to-lock (-38 degrees to 38 degrees) in about 1 second. The system can readily handle this.
We probed the torque sensor on the DeLorean.eu column a bit, and found that it was simply outputing a current proportional to the sensed torque on the handwheel. This made it easy to then implement our own power steering system, which really helped with the subsequent additions of faster steering ratio and front battery pack.
Upgraded Steering Rack
In Summer 2015, after replacing the stock lower control arm with our own A-arm design, and thereby removing the front anti-roll/stabilizer bar, we saw that we now had the space to greatly increase our steering angle with not much effort, by simply getting a faster rack ratio. The stock range, +/- 23 degrees, is really quite small. This would also make steering a bit more responsive for manual driving, as the original ratio was quite slow, possibly to make up for the lack of power steering. We found a really nicely built, fast rack from Woodward steering. To package it, we decided to remove the front crash member entirely, weld a bent steel doubler plate to the front cross member, and then use machined mounts to go from the rack to the doubler plate. This tabula rasa solution also makes any future work to the front area, i.e. the addition of the front pack, much easier. At this time, the new rack still interacted with the original upright/steering arm, and gave us about +/- 36 degrees of steering. Currently, with the new upright and steering arm, we are getting +/- 38 degrees of steering. Aaron Sellars was once again instrumental in providing suggestions and design reviews during this process, and for doing the bulk of the cutting and all of the welding - while I happily did the sanding and surface prep.
Front Battery Pack
The front battery pack, known internally as the 'frack', is installed in what used to be the front trunk, a.k.a. 'frunk'. This posed an interesting packaging challenge, with constraints from the steering rack, lower steering column, radiator + plumbing, and front suspension/wheel - especially with our relatively large steering range (+/- 38 degrees). We also had to decide what portions of the original fiberglass frunk to cut away - i.e. saving the major structure that holds up the bumper would avoid additional reinforcement work, and minimizing cutting the thick foam-filled bits would help reduce the general grossness of the operation.
A tubular steel subframe was designed and fabricated, and bolts to modified points on the rollcage and the crossmember; structural FEA helped drive and verify both subframe and mount designs. Tushar Goel and Joe Sunde, then a summer student in the lab, did the bulk of the design-for-manufacturing and fab work. Compared to the rear battery pack project 18 months previous, this felt like a much more mature design process and end product. Much less blind-leading-the-blind, and much more structured + scheduled planning, design reviews, and purposeful mentorship (hopefully Joe agrees with this statement). For fabrication, we discovered (were told by Aaron about) services that can laser-cut all the tubes from CAD weldments, including fishmouthing, mitering, and even tab-and-key to hold everything together, which was a total game changer.
The clean-sheet doubler plate we installed while doing the steering rack upgrade made the bottom mounts relatively straightforward. We used the steering rack mounts to temporaily bolt up the subframe, and designed + welded the upper mounts to the rollcage where the subframe lay. As that rollcage area is not really lined up to the planes of the car, we used an interesting 3d printed jig to measure the cage-bolt hole distances, and several rounds of cardboard/foamcore prototyping for validation. The final slotted-sheet mount was waterjet cut, ground down to fit the angles, and welded in place by Tushar. It was a bit of work, but sooooo satisfying when it fit!
Cooling had to be radically re-routed; getting a custom-sized radiator helped. Renovo once again helped immensely with design reviews, suggestions - and of course the actual kick-ass battery pack + new harnesses!
The interior was revamped in the whirlwind month leading up to MARTY's unveiling at the October 2015 Open Garage Talk hosted by Jamie Hyneman.
We worked closely with Renovo.auto, who took the lead on designing and executing the gorgeous bent sheet steel dash. The center console preserved much of the avionics-inspired functional minimalism I liked from the original prototype, which we ran for the first several months of operation. We added the `top bit', which houses a pair of Nexus 7 tablets, and also refined the overall design a little. The black ano finish, in particular, was an important learning - the California sun is bright and can cause big shiny panels to be quite blinding! The modular panels bolt to two main rails, and can be easily removed to access the wiring underneath.
The idea behing using Android tablets was to have a pretty open platform we could configure as needed. The current software on the tablets was designed by Renovo, with some design input from us, and shows off a bit of their onboard telemetry in action. The tablet panels were designed by Phill Giliver, and Mike Carter did most of the CNCing for the center console; I only ran the engraving.
2 years of use later, and the center console continues to be one of the mechanical design aspects of the project I am most proud of. The emergency stops and steer-by-wire/throttle-by-wire toggles are clearly and quickly accessed, and are all huge switches with nice beefy detents and authoritative CLICK noises. The drive selector goes R-N-D-Autonomous, with a positive mechanical interlock between D and autonomous - you have to lift the plunger to switch past D. I will be honest and say that when I first designed this, I had been motivated in no small part by the `coolness' factor. After much use, though, it turns out that this really does make operation a lot easier. When using the drive selector you don't have to worry about accidently going into autonomous mode, and so can naturally 'flick' it around like on a normal automatic PRNDL without second-guessing. There is a decent detent on the switching, which feels great; no big mechanical design went into this - I just sorted the rotary switches on Digikey by 'detent force' and picked the biggest one =P.
And, perhaps most importantly, I personally love the clean spacecraft-esque look and feel, which just makes testing feel that much more like the future...
The rollcage/seats/harness were designed and installed in Summer 2014, as part of the big electrification overhaul.
The rollcage ties to the front and rear strut towers, as well as the rear Y of the frame near the center tunnel. It is designed not only for safety, but to also add some much-needed torsional stiffness to the DeLorean. The gullwing doors, highly raked windscreen, and low roofline actually make this quite an interesting packaging problem. The roof longitudinal tubes need to run along the center to clear the doors, and the front hoop needs to tuck up into the windscreen pocket to avoid obscuring the already-quite-limited view of the driver. Removing the massively thick door trim made it possible to package the side tubes. The seats are mounted to a billet bracket that gives us adjustable tilt, which turns out to be really important for packaging taller people. The top two points of the six point harness tie to the rollcage, while the bottom four bolt through the fibreglass tub with aluminum backing plates.
Physical prototyping with PVC tube mock-ups really helped drive the design, and illuminate the above issues. Testing is most useful when you have a good set of test cases - this became quite clear the first time we let our advisor, Prof. Chris Gerdes, test out the PVC mockup. Being quite a bit taller than the rest of the team, his head stuck a full 3" above the roofline of the car, and the front roll hoop completely obscured his vision, even when crouched down. This drove the decision to have a tiltable seat and tucked front rollhoop.
Phill Giliver did the bulk of the prototyping/CAD design - and an awesome job at that - and we then handed it off to the amazing Tony at TC Design for both design review by a professional with tons of experience and the actual fabrication work.