Regenerative Braking is a way to recover braking energy and re-use some of it for acceleration later, for example in an EV or hybrid. It's a way to capture some of the otherwise wasted braking energy and make vehicles a bit more efficient. Note that regen has losses and can never be 100% efficient in round trip form, or we would have perpetual motion cars that never needed recharging. It's at best around 64% efficient overall. (Overall EV efficiency is about 80% so 64% for regen seems possible since 80% of 80% is 64%, and the battery, electronics and motor losses get applied twice: once for charging (braking) and once for discharging (accelerating)). Regen helps improve overall efficiency, but relatively marginally. That said, it's definitely worth having and a significant benefit of an EV or hybrid.
On electric vehicles and electric hybrids, regen typically slows the car by using its traction motor as a generator and putting the resulting energy back into the battery pack. The final part of slowing and stopping the EV or hybrid is done by conventional friction brakes which usually press an iron brake disc between brake pads. Friction brakes also come into play in an EV or hybrid when the need for heavy braking exceeds what the electric motor and battery can take back in as regen, but that much braking power usually isn't required in typical daily driving.
Friction brakes are wasteful since they turn the kinetic energy of motion into heat in order to slow the car. That heat is a total waste since it can do no further useful work. So it's desirable to maximize regenerative braking in an EV or hybrid and minimize friction braking.
However there is a serious error in the way regen is being controlled in some vehicles. A growing number of car makers put heavy regen on the throttle. This is a serious mistake. Putting heavy regenerative braking on lifting off the throttle violates some important principles of safety, ergonomics, human factors, driveability and vehicle dynamics.
Heavy regen on the throttle is wrong in terms of safety since it can exacerbate trailing throttle oversteer, a positive feedback condition that can result in loss of control and contribute to a crash. During oversteer, the tires at the rear of the car lose traction before the fronts. This can happen under extreme limit conditions such as an emergency while driving on the street or much more commonly during the rigors of racing. NASCAR racers call this "getting loose", and describe it as the back end of the car "coming around." You may know it simply as a "spin," since it's a common way for cars to "spin out," with the rear of the car spinning around the front.
Trailing throttle oversteer is most commonly initiated by lifting off the throttle or braking abruptly in a turn. It can also begin while driving in a straight line, for example when slowing suddenly to avoid an obstacle in front of the vehicle. Lifting off the throttle or braking abruptly transfers load rapidly from the rear of the car to the front due to the mass of the vehicle acting as a lever through the front and rear suspensions. This upsets the balance between the traction at the front and rear of the car because it reduces the vertical load on the rear tires while increasing it on the front. If the car is yawing (turning) at the same time, the decrease in load on the rear tires reduces their traction, and if they exceed their friction limits, the back end of the car can "step out" and start to spin from the rear.
Most non-racing-trained drivers don't know how to react to trailing throttle oversteer: the correct response, after losing some speed to gain back some margin of traction, is to ease gently back onto the throttle to transfer load back onto the rear tires to stop the oversteer. This is exactly the opposite of what most untrained drivers usually do. Instead they will often lift further off the throttle and make the oversteer worse, probably leading to a loss of control of the vehicle and a possible crash. Lifting off the throttle transfers load off the rear tires and onto the fronts, which is usually the opposite of what's needed. Reduced vertical load decreases traction, and having less traction at the rear makes the spin worse.
Trailing throttle oversteer is a positive feedback loop because lifting off the throttle often starts oversteer, and lifting even more further increases oversteer. Positive feedback in this case is a bad thing since it can lead to increasing oversteer and a loss of control.
Key here is that abrupt braking or heavy regen can cause a vehicle to oversteer or make ongoing oversteer more severe. Coupled with the natural reaction of untrained drivers to lift off the throttle when they start to lose control, putting heavy regenerative braking on the throttle is exactly the wrong thing to do in the context of trailing throttle oversteer. It can be equivalent to hard braking, which is essentially the worst thing to do during oversteer. Therefore putting heavy regen on the throttle is a significant ergonomic and safety error that may contribute to crashes.
Oversteer has a counterpart known as understeer. Understeer is when the front tires lose traction before the rears during a turn, and dirt track racers call it "pushing." The car goes straight instead of following the curved path you're steering, or traces a wider than desired path. Understeer, in contrast with oversteer, is inherently safer because it's a negative feedback condition; understeer tends to be self-limiting in its most commonly occurring forms. Lifting off the throttle of an understeering car slows the car, pulls it back from the limits of adhesion, and transfers some load back onto the front tires. The added load increases the front tires' traction and decreases the understeer. This is a negative feedback loop, and it's desirable because it's mostly self-correcting.
Most production cars are tuned to understeer because of this. Front-wheel drive cars also tend to understeer because the front wheels lose traction before the rears since they're essentially asked to do too much of the braking, turning and accelerating. (Rear-drive cars divide up the work more evenly between the front and rear of the car.) Desirable understeer for untrained drivers is also one reason for the prevalence of front-wheel drive, in addition to other factors such as packaging efficiency and simpler manufacturing. Porsche was famous for tuning understeer on its rear-heavy 911s in order to compensate for their tendency to oversteer. (911s are too rear-heavy because the engine is behind the rear axle; this is a design that tends to oversteer. All major, modern race cars such as Formula One are mid-engined where the engine is in front of the rear axle, which is optimal weight distribution.)
BTW any car can be driven into understeer or oversteer, intentionally or otherwise. Getting simultaneous oversteer and understeer is also possible; it's called a 4-wheel drift, where a car drifts out sideways in a turn. 4-wheel drifts are the prominent feature in the sport of drifting.
Heavy regen on the throttle is wrong in terms of ergonomics because it cross couples braking control with acceleration control. These are separate functions, and the controls should reflect that. Internal combustion engine vehicles and those with automatic transmissions have some drag when lifting off the throttle, but this unintentional side effect is generally mild and usefully helps to gradually slow down the car, such as when coming to a highway off-ramp or intersection. In contrast, heavy regen on the throttle feels like lifting off the throttle in a manual transmission car in a low gear, and it's not something ordinary drivers would normally do. It's unnatural and generally undesirable since it can lead to a loss of control if extreme enough.
Tesla takes some steps to mitigate this, for example by using progressively lighter regen as speeds increase (personal correspondence with Tesla founder Martin Eberhard), but it doesn't diminish the concept that heavy regen shouldn't be on the throttle to begin with. It's still a mistake. The bulk of regenerative braking should be on the brake pedal, where its function is normally and intuitively expected.
Unfortunately heavy regen on throttle lift has become a fad in EVs. It may very well have been a convenience when it showed up in friend-of-a-friend Alan Cocconi's early EVs, in order to not need to modify the stock braking systems of the internal combustion cars they were based on. Tesla appears to have inherited this from Cocconi's AC Propulsion technology, and like the emperor's new clothes, other automakers have blindly copied Tesla, perhaps thinking it was somehow fashionable. If so then this is a good example of the dangers of groupthink. Tesla and others should follow Elon Musk's own advice and reason this from first principles, particularly from ergonomic and safety principles (and very basic vehicle dynamics), and put regenerative braking on the brake pedal where it properly belongs.
Toyota, GM and others correctly put a majority of regen on the brake pedal, and they have decades of human factors science to support it. They also have informed common sense, taking into consideration the untrained drivers' reaction to trailing throttle oversteer. It's true that traction control and computer overrides of vehicle dynamics can compensate for this to a limited extent, but they can't overcome the laws of physics. If the vehicle's controls are properly implemented, they might not need to.
Multiple generations of Toyota and GM EVs and hybrids including the GM EV1 and Toyota Prius and original RAV4 EV have correctly put regenerative braking on the brake pedal. The first generation Prius regenerative braking implementation was a bit clunky, but later generations of Prius, the original RAV4 EV and GM's spectacular EV1 have smoothly integrated regenerative braking that progressively blends with friction braking as brake pedal force is increased. This is difficult to do well and is ergonomically highly desirable because most of the braking function happens on the brake pedal, as a driver might expect and want.
A valid argument can be made that this arrangement is slightly less efficient because a slight amount of friction braking may be used at light application of brakes, but the electric traction motors are powerful enough that significant power levels go into the motor and back into the battery pack before heavy use of friction takes effect at higher levels of braking, Certainly a large portion of relatively mild braking most drivers use in daily driving can be recaptured by regen on the brake pedal. Only heavier application of the brakes causes the balance to shift toward the friction brakes. In other words, most people don't brake hard enough in street driving for the difference in efficiency to be large. A large majority of brake energy in ordinary driving will be captured by regen since most drivers don't exceed the braking power of the motor even when the regen is integrated with friction brakes. This is commonsensical given the relatively gentle way most people drive and the relatively high power of traction motors.
In a sense, Tesla also proves this point with their heavy use of regen to slow vehicles in daily driving and only marginal use of the friction brake to come to the final stop. Even though their regen functions only on the throttle and is not integrated with the friction brake system, regen does most of the braking. The same is largely true of EVs and hybrids where the regen is properly integrated with the friction brakes: most of the braking is done by the motor. Differences in efficiency between regen on throttle and regen on brake are likely minor compared to the ergonomic and safety benefits of the latter.
One of the weaker arguments in favor of regen on throttle that sometimes comes up when discussing Teslas in particular is that regen is best on the front axle, so real-wheel drive cars that don't have a front motor need to brake on throttle lift in order to capture more of the energy that would normally go to the front brakes and be wasted as heat there. While it's true that most braking power belongs on the front axle since load transfers to the front during braking, in daily driving this is largely a non-issue since the typical braking power is so low. It's much more of an issue during high-power, limit-condition braking during racing. It's also not really an issue during a panic stop on the street either, since most of the very-high-power panic braking would be handled by the friction brakes, even on EVs and hybrids.
Street and race cars both have brake power carefully balanced between front and rear. Conventionally this is done by tuning hydraulic brake piston diameters and brake pad area on the brake rotors (or brake shoes in drums) so that when the brakes are applied, more braking force is applied to the front wheeels. It's part of the reason why the front brakes and pads are larger than the rears. (The other major reason is heat capacity; the front brakes do more work so they need to be able to absorb and dissipate more heat. Relative work and heat dissipation are also reasons why front brake rotors are more often vented than rears.)But again most of the above only really comes into play during racing. It's probably not legal to drive hard enough on the street for the brakes to be worked that hard, though high brake loads could happen during a panic stop. Electric motors have more than enough regen power to do most of the braking of street driving whether the electric motor is on the front axle or rear. In practical terms, there are two modes of braking on the street: gentle normal driving which can be easily handled by either front or rear electric motors, or panic stops which are largely handled by the friction brakes.
Most mass production EVs such as the Leaf, Bolt and others are front-wheel drive for various reasons mentioned above. Even on those, the regen belongs on the brake pedal for ergonomic and safety reasons. Front-wheel drive cars do have the advantage of having the motor on the right wheels to make the best theoretical use of regen, but again typical street driving won't exceed the regenerative braking power of either front or rear electric motors, so it doesn't make a positive argment for regen on throttle. On the contrary, real world experience with mild street braking argues against it, especially when taking into account safety and ergonomic considerations. Since normal brake power usage is so low, most of it can be handled by regen even on a rear-motor-only EV (as Tesla has shown), and that rear-motor regen can and should be integrated with the brake pedal instead of the throttle.
A majority of Tesla drivers may disagree with the above, but most of their experience will be informed by a very limited set of driving experiences typical of most street use, and not racing, limit behaviors or emergency maneuvers. They may like the emperor's new clothes, blissfully thinking the new "clothes" are really cool, when in fact they're a naked ergonomic error.
Most drivers also use only a tiny portion of the performance envelope of their cars, and given an unfortunate lack of training, that's probably appropriate. If they have occasionally approached some friction limits of a car, most often they will have their experience distorted by electronic traction controls, which blunt the responses during limit conditions, masking the true underlying vehicle behavior, up to a point. Beyond that point, the laws of physics always win, and heavy regen on throttle lift leading to oversteer is one way to exceed the limits without fundamental understanding of why it happens or how to correct it when does. In contrast, racers and vehicle dynamics engineers understand exactly why this issue is important.
I happened to have a long talk with a Tesla engineer and brought up this issue with him. He mentioned that there were two camps of thought about it inside Tesla. Some agree that heavy regen belongs on the brake pedal and others feel it belongs on the throttle. He pointed out research that shows heavy regen on the throttle led to more efficient energy use. (I can't recall if he said this was based on simulation or actual driving.) That's possible, given the lack of control sensitivity by most street drivers compared to good race drivers, but it doesn't change the safety and ergonomic issues above which should take precedence over some possibly minor differences in efficiency.
More efficient driving can also be trained interactively with the car, as Nissan showed with its growing tree dashboard animation in the Leaf. To Tesla's credit, their energy usage histogram display does encourage smoother, more efficient driving, if it's used and interpreted correctly. For maximum efficiency, energy usage should be low and smooth. That will happen with smooth and gentle application of the controls. When regen is correctly integrated on the brake pedal, gentle application of the brakes will maximize the use of regen to slow the car and minimize the use of wasteful friction braking. (BTW the fastest racing drivers are often the smoothest. Watch Lewis Hamilton's hands on the steering wheel of his Formula One cars some time; they move smoothly. The reason is that at the extreme limits of adhesion, even the smallest control disturbance can push the car over those limits.)
To be clear, the problems with heavy regen on throttle may not be apparent in daily driving. It becomes the most critical in emergency or limit conditions, where it may cause an undertrained driver to do things that will lead to loss of control due to oversteer and a possible crash. It's incorrect in terms of basic human factors science and first principles of vehicle dynamics. If safety truly matters, heavy regen on throttle is a poor idea that should be banished. It's likely that the emperor's new clothes and groupthink at least partially underlie the preference for it. The case for marginal improvements in theoretical efficiency is weak compared to the much stronger and deeper knowledge we have about safety, human factors and vehicle dynamics. Having regen maximized through the brake pedal's initial range of application facilitates maximum energy recovery while simultaneously being ergonomically correct safety engineering.