# What determines when an automatic transmission shifts?

I'm working on developing an automatic bicycle (mostly microcontroller based) and I'd presume I'd be shifting in a similar way to a car. So my question is, what factors does the transmission take into consideration when shifting? I know when I drive my car, it tries to shift at a constant engine rpm, but when I accelerate fast, say when entering the highway on a short entrance ramp, the tranny doesn't exactly keep up with the engine rmp, which may be at 5 or 60000 rpm before shifting.

So is it concerned more with speed, rpm, some other factors,or all of them? Which ones take precedence in its decision to shift?

So for instance, my main concern is riding the bike downhill. Now if you're riding a regular bike like me, you would stop pedaling and let gravity do the work for you. Does an automatic transmission do this as well or does it shift up to match the speed or down to match the low rpm?

• In a legacy hydraulic automatic, it's determined by one of these: mechanics.stackexchange.com/a/2794/57 – Bob Cross May 2 '14 at 17:00
• @BobCross ... this question should most likely be covered by your linked answer. Consider closing as duplicate? I like your use of the term "fuzzy logic" as that covers it pretty darn good ;-) – Pᴀᴜʟsᴛᴇʀ2 May 2 '14 at 17:05
• @Paulster2, not necessarily - if audiFanatic is building a computer-controlled transmission for a pedal bicycle, the hydraulics are interesting but not sufficient. – Bob Cross May 2 '14 at 17:23
• @Bob_Cross Yes, I'm not really concerned with how old analog transmissions work (though it is interesting). I'm much more interested in more modern, electronic-based transmissions. – audiFanatic May 2 '14 at 17:39
• Or rather I'm interested more in the logic behind shifting rather than the mechanics in how it works. My system is servo-based, so the mechanics are completely different. – audiFanatic May 2 '14 at 17:42

## 3 Answers

I'm working on developing an automatic bicycle (mostly microcontroller based) and I'd presume I'd be shifting in a similar way to a car.

Well, you're in for a shock: it's going to be pretty different.

So my question is, what factors does the transmission take into consideration when shifting? ... So is it concerned more with speed, rpm, some other factors,or all of them? Which ones take precedence in its decision to shift?

It depends. The ruleset is going to be situation dependent and somewhat non-deterministic (also known as using fuzzy logic). A car ruleset is also going to be very different from one that you'd use on a bicycle.

Here are some examples of the conflicting priorities that you face in a car's automatic transmission:

1. Fuel economy
2. Acceleration
3. Throttle response (is there a shifting delay between stomping on the gas and feeling the go?)
4. Fuel economy (what, again?)
5. Annoyance (oscillating shifting is annoying: e.g., 3-4-3-4-3-4).
6. Fuel economy (yes, yes, we get it!)

The point being, the transmission is trying to optimize a bunch of conflicting metrics looking for a local maxima.

So for instance, my main concern is riding the bike downhill. Now if you're riding a regular bike like me, you would stop pedaling and let gravity do the work for you. Does an automatic transmission do this as well or does it shift up to match the speed or down to match the low rpm?

On a bicycle, you have a complete different set of problems. For example, I like to pedal at a fixed pace and I'll fiddle with the gears regularly to match my road speed to my pedaling pace. Your automatic transmission would need to gear up and down according to road speed. It would also need to anticipate the upcoming hill - I do, since I can see the slope and gear down. Waiting until my pace slows down will be annoying. It would be better to detect the situation (e.g., using a level or GPS data).

On a downhill slope, I gear up so that I can build up momentum. I also like knowing that I'm in the right gear for my road speed even if I'm coasting for a bit: I don't like the overspeed situation where I can end up running over my own foot! In an automobile transmission, the engine almost decouples in a coasting situation (even though the engine is still "pedaling").

Finally, an automatic bicycle transmission can't oscillate between gears. In a car, it's annoying to hear the engine jump from 2000 RPMS to 3000 RPMS and back. On a bicycle, where there's a distinct shock to the system with every shift, the rider would quickly take a hammer to the transmission if it were switching every second or so.

• Excellent, very clear answer. I can't say I'm surprised, but I really thought there would be more similarities. Regardless, the compare/contrast you laid out is very helpful. I'm currently using a gyroscope to determine when I'm approaching a hill, so that should be sufficient (I just wish the manufacturer was more helpful and provided a library, but that's a side issue). I'll leave the question open for a few hours to give others a shot, then I'll select a best answer; most likely yours. – audiFanatic May 2 '14 at 17:52
• @audiFanatic, feel free to swing by the chat - I'd be happy to discuss issues further. It's a good idea, it's just a super hard problem to solve in a way that a human is going to think is better. – Bob Cross May 3 '14 at 1:09
• Yes, I'm finding that out the hard way with this project, I definitely bit off more than I could swallow with one. On the flip-side I guess I have one of the more interesting and complex senior design projects, at least from an electrical and computer engineering standpoint; a mechie on the team would have been nice. :D – audiFanatic May 3 '14 at 1:32

I believe the @bobcross answer is pretty good, but wanted to add some more information which I believe is pertinent to the question, yet does not negate what Bob has written:

You asked about the logic behind how an automatic transmission shifts and why it shifts when it does. To answer part of this, we need to look at what actually controls the transmission. This unit (in newer automatic vehicles) is called the transmission control unit (TCU -- obviously, lol). It is a unit separate from the engine control unit (ECU), but does work in conjunction with it. The inputs used for it to "decide" when to shift are as follows (taken from Wikipedia):

Vehicle speed sensor (VSS) - This sensor sends a varying frequency signal to the TCU to determine the current speed of the vehicle.

Wheel speed sensor (WSS) - Modern automatic transmissions also have a wheel speed sensor input to determine the true speed of the vehicle to determine whether the vehicle is going downhill or uphill and also adapt gear changes according to road speeds, and also whether to decouple the torque converter at a standstill to improve fuel consumption and reduce load on running gear.

Throttle position sensor (TPS) - The TPS sensor along with the vehicle speed sensor are the two main inputs for most TCUs. Older transmissions use this to determine engine load, with the introduction of drive-by-wire technology, this is often a shared input between the ECU and TCU. The input is used to determine the optimum time and characteristics for a gear change according to load on the engine. The rate of change is used to determine whether a downshift is appropriate for overtaking, for example, the value of the TPS is also continually monitored during the journey and shift programmes are changed accordingly (economy, sport mode, etc.). The TCU can also reference this information with the vehicle speed sensor to determine vehicle acceleration and compare this with a nominal value; if the actual value is much higher or lower (such as driving uphill or towing a trailer) the transmission will change its gearshift patterns to suit the situation.

Turbine speed sensor (TSS) - This may also be known as an input speed sensor (ISS). This sensor sends a varying frequency signal to the TCU to determine the current rotational speed of the input shaft or torque converter. The TCU uses the input shaft speed to determine slippage across the torque converter and potentially to determine the rate of slippage across the bands and clutches. This information is vital to regulate the application of the torque converter lock-up clutch smoothly and effectively.

Kick down switch - One of the most common inputs into a TCU is the kick down switch which is used to determine if the accelerator pedal has been depressed past full throttle.[3] When activated the transmission downshifts into the lowest permissible gear based on current road speed to use the full power reserves of the engine. This is still present in most transmissions though is no longer as necessary to use as the TCU uses the throttle position sensor and rapid rate of change to determine whether a downshift may be necessary, thus there is no need to use the kickdown feature in most circumstances.

Brake light switch - In more modern TCUs this input is used to determine whether to downshift the transmission to increase engine braking effect if the transmission detects that the vehicle is going downhill.

Traction Control System (TCS) - Many TCUs now have an input from the vehicles traction control system. If the TCS detects unfavorable road conditions, a signal is sent to the TCU. The TCU can modify shift programs by upshifting early, eliminating the torque converter lock-up clutch application, and also eliminating the first gear totally and pulling off in 2nd.

The TCU uses these inputs, cross references them against tables within its memory stores, and decides which gear it should be in. As it states in the Wiki article, modern TCUs are so complex in their design and make calculations based on so many parameters that there are an indefinite amount of possible shift behaviors.

Well - most older cars that didn't have a TCU used manifold vacuum to determine shifts. When one of the vacuum hoses broke or fell off, the transmission wouldn't shift properly.

But this won't be realistic for your bicycle - you'll have to write some software and figure out an algorithm that will work - using some speed and torque sensors.

It's new uncharted territory!