Reflections on the potential of human power for transportation

Saturday, October 10, 2015

GM Lean Machine: The Solution to Personal Mobility, 30 Years Before the Competition

The problem is obvious to even the most causal driver sitting in traffic. Most cars have just one occupant, and for that occupant the car is much larger than it needs to be; at least twice as wide, equally long and maybe 10 times as heavy. As a result those cars consume more fuel and take up more space than is necessary.

A potential solution, while not as obvious, is rather straightforward. Make the vehicle just wide enough to seat one driver. Vehicles of this type are known as man-wide cars or narrow-lane vehicles. For two people, seat the second person behind the first in tandem. Keep the driver’s head height at sport-car level and to prevent the vehicle tipping over in corners, allow it to lean like a bicycle.

Both Toyota with its i-Road three-wheeler and Nissan with its four-wheel Land glider have recently built such vehicles.

But probably the most surprising fact is that 33 years ago, General Motors built such a vehicle. It was conceived by Frank Winchell, GM’s V.P. of engineering. With over 50 patents, Winchell was a strong creative force at G. M. He also assisted Jim Hall of Chaparral racing fame by designing and building many of the components that went into their cars. But Winchell’s most revolutionary contribution must be the Lean Machine.  GM said “it may be the first new road vehicle invented this (the 20th) century”. The Lean Machine took seven years to develop. That means work began on it shortly after the gas shortage of the early 1970s. Just as impressive as the visionary idea that such a vehicle was necessary is the design execution. It is probably as simple and as elegant an implementation of a motorized leaning vehicle as could be designed.

Ultimately the Lean Machine ended up in a GM World of Motion exhibit at Epcot. The images below are from the publicity brochure for the vehicle.

To be accurate, the Lean Machine is only a semi-leaning vehicle. The motor module, with its two driving wheels, does not lean. Only the passenger pod and the front fork lean. But, since the majority of the weight is in the motor module and since the center-of-gravity (cg) of that module is low, the overall effect allows for 1.2g cornering with 50deg. of lean.
George Georgiev borrowed heavily from this design for his leaning Varna Cargo Trike.

Paraphrasing the brochure, wheelbase is 71”, track is 28”, length is 122”, width is 36” and height is 48”. Actual vehicle weight was 350#. 

The chassis was made up of two parts. The rear section was made up of a tee-shaped frame. A rectangular tube that ran transverse to the vehicle supported the engine and the swing arms for the suspension. Attached to the rectangular tube was a horizontal spine made of a large-diameter tube. This spine ran from the back to the front of the vehicle. It was to this tube that the leaning portion of the chassis was attached

The front portion of the chassis was made up of a wishbone or Y-shaped frame made up of round and square tubing. The two legs of the Y supported the seat. The single tube portion of the Y held the steering head of the fork.  This tube also had a vertical extension that held a spherical-bearing rod end. That rod end was bolted to the end of the horizontal spine of the rear chassis. The front chassis was also connected to the spine by a sleeve bearing behind the seat. These two bearing allowed the seat and fork to rotate about the spine which resulted in the leaning action.


Leaning was accomplished by pushing on either of two levers with the driver’s feet. The levers were attached to cables that wrapped around the spine. Pushing a pedal unwraps its cable and rotates the cabin away from that pedal. At the same time the cable for the other pedal is wound up. The picture below shows a precursor to the LM. Notice the vehicle is leaning away from the driver’s extended leg.

When stopping, the driver was required to hold the passenger pod upright with his feet until a lever could be engaged to hold the vehicle upright. If the driver forgot, the pod would flip sideways, but LM did not tip over. This is a benefit of the semi-lean approach where the motor module’s weight keep the vehicle upright when this happened.

Below is a short clip of the Lean Machine in action.

To keep the nose-region low, Showa forks from a small displacement motorbike were used supporting a 3.00-10 mini-bike tire. Steering controls were 18” behind the fork, pivoting about a horizontal axis and connected to the fork through a linkage.
The engine used was from an 185cc Honda ATV with a 5-speed semi-automatic transmission driving the rear wheels through a garden-tractor differential. The engine put out between 12 and 15hp. The rear tires were 4.80-8 boat-trailer tires. The engine required a cable-pull starter which was accessed through a hatch in the motor-module cover.

The cd for the vehicle (coefficient of drag) was .35, not exceptional but the vehicle cross-section was small, so aero drag was low. 

Performance numbers were a top speed of 80mph and a fuel economy of 120mpg at a steady 40mph.
Notice that the 120mpg number does not match the 200mpg value published in the Epcot brochure. The 200mpg number is theoretical, based on a 38hp engine and a more streamlined body.

That body is shown in the mockup shown below. Notice the lack of the gap between the passenger pod and the motor module that would exist in a working vehicle. This more streamlined version had a cd of .15.  One of the reasons for this lower value is the tapered rear of the passenger pod as opposed to the flat-back rear of the actual vehicle. The line drawings in the brochure are of this tapered version.

The last model is of a flying-car version of the Lean Machine. While it looks very cool, the presence of even folded wings defeats the narrow-width paradigm of the original design.

Of course, if resurrected today, one change would be to have the leaning be computer controlled. The driver would just steer. Since leaning is a function of turn radius, vehicle speed and, in this case, weight of the occupant, the use of accelerometers and strain-gauges feeding information to a micro-processor which in turn controlled hydraulic actuators would take the guess work out of leaning.

The LM, as configured was just a one-passenger vehicle. It would be easy enough to extend the passenger pod to hold a second seat. The second seat location could also serve as a cargo compartment. And like the Messerschmitt tandem cars of the 1950’s, if the cg is located at the rear seat, the addition of a passenger or cargo would not affect the vehicle handling. More weight in the passenger pod would require less of a lean angle for a turn of a given acceleration.

Finally, the low-slung motor pod would be an ideal location to house heavy batteries for an electric vehicle. 

The technical information for this post came from a January 1983 Road and Track article by Peter Egan. The chassis photos from the Epcot exhibit came from Max Hall’s Maxmatic website.

Friday, September 4, 2015

Make My Velomobile a Quad!

I was talking to my son Kyle the other night. I told him I had finished the testing of my leaning-trike proof-of-concept commuter vehicle and that I had met the cruise-speed goals I had in mind for the project. I also told him about the latest lean-lock mechanism I had incorporated to ease stopping and starting with its faring in place. He listened patiently and after I have finished explaining, he pointed out that this was to be a commuter vehicle and the requirement to lock and unlock the leaning for stopping and starting was too complicated to expect riders to have to cope with. And I couldn’t argue with him. 

Now Kyle says he doesn’t read my blog, too much information, even though he is a road cyclist and an accomplished mountain biker. He has had the misfortune of riding and crashing one of my ill-fated rear-steering leaning-trike prototypes. With misplaced fatherly devotion, I was more concerned with the bent-up trike than his welfare. So he has a right to mistrust his old man’s hair-brained contraptions
Sometime after that I was reading Miles Kingsbury’s account of his riding his Quattro on the Roll Across America velomobile tour and how his four-wheel vehicle handled road imperfections better than the trikes with whom he was riding. Below is another picture of the Quattro in its convertible guise.

And another with the top in place.

So I had to ask myself, am I avoiding a four-wheel layout for my human-powered commuter vehicle because it is too simple to satisfy the gizmologist in me? Since I have no plans to use powered assist, the vehicle-classification reasons for not using four wheels is not an issue. So rethinking my assumptions is in order.

The historically significant velomobile, the Pedicar had four wheels. However its 38” width made it difficult to fit thru some barriers and its exposed wheels gave it poor aerodynamics. And of course it had the linear-pedal drive that had some efficiency issues.

It probably is not a good starting point to speculate on an efficient quad velomobile design.

The Quattro, on the other hand, IMO is too close to the ground for commuting in traffic. I do like the fact that the steering wheels are enclosed. Driving and steering the front wheels does make things complicated. There are the velocity fluctuations associated with single-universal joints. There does not appear to be any differential to account for the different wheel speeds during cornering.

Another candidate the design of a commuter quad is the Quest velomobile. It also has enclosed steering wheels and I have always been impressed by the level of refinement in its implementation. And its overall width of 30” is narrow enough to fit through bike-trail barriers.

Now my assumption is that the Quest’s current rollerover resistance is adequate for most situations. That being the case, the addition of a fourth wheel should increase that resistance by about 1.5X. Since rollover resistance is a function of vehicle track (spacing between the paired wheels) divided by the c.g. height, one could use that 1.5X to increase the c.g. height and maintain the three-wheel rollover resistance.

Referring to the dimensioned drawings, 

the faring behind the rider’s head is about 34” off the ground. That puts the riders head about 36” off the ground. If this could be raised to 48”, I would feel a lot more confident riding the vehicle in traffic. If the c.g. is raised 48/36 of 1.33 times, this is within the 1.5X rollover margin that the fourth wheel contributes.

Raising the rider and making the rear faring wider to accommodate the fourth wheel will make the vehicle less aerodynamic, but this should be acceptable considering that the original vehicle had a racing heritage and a commuter does not need to be as fast. One benefit to raising the rider's position is that ease of entry and exit can be greatly improved.

Since the rider is up higher, the steering can be located beneath or alongside the seat. This will allow the faring to be lower between the rider’s knees and his head and provide better forward visibility,
So the front end construction of the Quest quad can remain the same.

Behind the seat, the drive chain has to drive two wheels instead of one. The easiest way to do this is attach the cassette to an intermediary jack shaft. At the ends of the jackshaft, at the same width as the rear wheels, attach single-speed freewheels. Then connect the wheels to the jackshaft freewheels with two additional chains. The two-freewheel drive approach produces a posi-traction-like drive. That in combination with at least 50% of the vehicle weight being on the drive wheels (as opposed to 33% with a tadpole trike) provides exceptional traction for slippery-road conditions.

Depending on which side the rear wheels are being driven from and how the freewheels are threaded, one of the freewheel ratchets, its threading and the cog threaded on one of the rear wheels may need to be left handed. Left handed freewheels are available as BMX components.  The rear-wheel suspension can be based on the wheels pivoting around the jackshaft, so chain tension is not affected by suspension movement.

Yea, the quad conversion may not be that simple as I just proposed, but the Quest Quad would make a very attractive commuter vehicle.


P.S. Thanks to long time friend and HPV'er Jerry Onufer for this link to a very well executed cycle car, the Podride.

Tuesday, September 1, 2015

Trikes & Leaning Trikes: A Second Look

Why designers select three wheels for vehicles:

Compared to their four-wheel cousins, you see very few three wheel automobiles. I think that the basic reason for this is that, for layouts where all three wheels are equally loaded, and the vehicle tracks (width over the paired wheels) are the same, quad wheelers can corner at a 50% higher accelerations than trikes. Thus trikes need to be about 50% wider than quad wheelers for comparable tipping resistance.

So why is there continued interest in trikes by progressive vehicle designers? One reason it that tricycles are less regulated that quads. If light enough (less than 1500 lb. in the US?) the vehicles are considered motorcycles. And if the power is low enough, they can be considered mopeds. And if slow enough (less than 20mph in the US) they can be considered electric bicycles and have access to bicycle paths.

Another reason trikes are popular platform for new vehicle concepts is that they can be simpler than quads. For tadpole layouts (two wheels front, one behind) they can use motorcycle or bicycle transmissions to drive the rear wheel. For delta layouts (one wheel front, two behind) they can use motorcycle or bicycle fork steering. And delta trikes can be more aerodynamic than quads.

A three-wheel pedal-electric vehicle:

 So let’s look at a hypothetical pedal-powered commuter vehicle with electric assist. We make it a three-wheeler to take advantage of the reduced vehicle regulations. We want car-type seating with pedals about 10” below the seat, so we choose the tadpole layout. The pedals easily fit between the two front wheels.

We want a rollover resistance of at least one gee. The rollover gee limit for static vehicles is approx.


Where w is the track width (distance between paired wheels) and hcg is the height of the center-of-gravity off the ground. For tricycles, where the weight is equally distributed over the three wheels, w is approx. 2/3 the spacing between the wheels.

Automobiles regularly have rollover resistances in excess of one gee. And some sports cars can approach two gees or more.

For the one gee roll over resistance and a hcg of 16", we will need a track of about 48".
Both the Elf from Organic Transit and the Velocar from VeloMetro are actual examples of our hypothetical tricycle layout vehicle. Not surprisingly, they both have tracks of 48"

The Elf is shown below.

The proof-of-concept for the Velocar is shown below.

Does the rollover resistance need to be as large as one gee? Judge for yourself. The vehicle below is a Mango velomobile,  one of the better designed tadpole trikes. The vehicle width is 30” and I estimate the track to be about 28”. If hcg is 13”, Gees= approx. 0.7. Clearly in the case below it is not high enough for the turn that is attempted. And the turkey isn’t even wearing a helmet…

If a track of 48” is too wide for the paths the vehicle is to use, there is an unconventional trike layout that can be employed. I call it the Coventry-trike layout in honor of James Starley’s Coventry Rotary tricycle of the 1880s.

A modern version of the design would probably use equal sized wheels, all cantilevered for ease of removal and tire replacement.  One advantage of this approach is that, if the drive wheel is located along the c.g. in the side-to-side direction, 50% of the vehicle weight in on the drive wheel as opposed to the conventional tadpole layout which only has about 33%. 

One downside of this layout is that the trike would tip forward during hard braking. Tipping could be prevented, however, by having a caster near the ground opposite the front steering wheel and in front of the driving wheel. This caster could be frictionally loaded so when the trike tipped forward and the castor contacted the ground it contributed to braking.

The most complicated method of reducing the width of the trike, but maintaining its rollover resistance is to allow it to lean. The designers at VeloMetro considered making their Velocar a leaning trike, but since their business model was to rent their trikes to people for short trips, the associated learning curve would have been impractical. Leaning trikes have all the issues of recumbent bicycles, plus more if the vehicle is in and enclosed body. See below,

For bicycles and free-leaning trikes, 

                Gees=tan (alpha)

Where alpha is the lean angle from vertical. For one gee of rollover resistance the lean angle is 45deg, which is quite a bit of leaning, and is to the state when the tire contact with the ground could be on the sidewalls, not the best for adhesion.

If properly implemented, however, leaning trikes can develop more rollover resistance than just the bicycle-lean effect.

Since I have spent a good deal of time working with the delta trike layout, let us examine another hypothetical vehicle. It has a width of 30” and a track of 26”, a wheelbase of 48” and a c.g. height of 24”. It can lean 30deg. from vertical.

Below are diagrams of this trike seen from the top and the back. The trike is fully leaned over.

There are four rollover accelerations associated with this design. Two are common and the other two are associated with leaning trikes.

(Lean locks are discussed in more depth in the post above.)

The first acceleration is when the trike is locked (lean-locked) in its upright position. This upright–locked acceleration is 8.5” (dim a. in figure 1. from the upright c.g. to the tipping line A-B) divided by 24” (the c.g. height) or .35gees.

The second acceleration is due to leaning (like a bicycle). This free-leaning acceleration is tan(30deg) or .58 gees.

A leaning trike can corner harder than a bicycle for a given lean angle because the wheel opposite the lean direction increases the distance from the c.g. to the line of tipping This acceleration is when the trike is fully leaned over and the lean-lock is engaged. This lean-plus acceleration is 20” (dim b in fig.1 the distance from the c.g. to the tipping line A-B) divided by 20.78 (dim d in figure 2). The lean-plus acceleration is .96 gees, almost .4 gees greater than the free-leaning acceleration.

The last acceleration is associated with the tendency of the trike in the leaned-over but lean-locked state to tip in the opposite direction. This counter-tipping acceleration is 3.5” (dim c in fig 1, the distance the c.g. overhangs the tipping line A-C) divided by 20.78 (dim d in figure 2) or .17 gees.
Sounds confusing, doesn’t it.

To try and make things clearer, let’s have our rider, from the Mango tipping video above, ride our trike through a progressively tightening turn. The rider gets into the delta trike with leaning locked upright. The rider begins to make a turn. When the acceleration of the turn reaches .35 gees, the lean lock must be released to prevent tipping. The turn gets tighter and the rider leans into the turn. As the turn gets to .58 gees the rider is at the lean limit. If the turn gets tighter the rider will not be able to maintain the lean and will be pulled upright. However, if the lean lock is again engaged at maximum lean, the rider can corner with an extra .38 gees without tipping.

Now even though the trike is locked up with the c.g. outboard of the left tip line (A-C) the trike won’t tip over as long as the turn is at least .17 gees, the counter-tipping acceleration. The acceleration of the turn must drop from the .96 gee max to .17 gees before the lean-lock needs to be released and free leaning can resume. This is plenty of time to do this.

The requirement to continually engage and disengage the lean lock may be too much of a complication for a human-powered commuter vehicle. But for powered vehicles that can employ automatic lean actuators based on vehicle accelerations, this would all be transparent to the user. The driver would just steer and the onboard computer would determine the appropriate amount of lean.