Test-riding the self-driving Porsche, a computer that carves curves

Porsche is currently the only company in North America to offer a rear-engine, rear wheel-drive car (let's debate the Smart Fortwo later). It is also the only sport/luxury manufacturer to offer a trio of plug-in hybrids in its current lineup — the Panamera, Cayenne, and 918. And it is (subjectively) the only brand to successfully deliver exquisite road feel with an electrically assisted system. So it should come as no surprise that the stalwarts from Stuttgart should scythe their own path toward an autonomous future.

In fact, the engineers working on this pre-production system, known internally as InnoDrive, insist that it isn’t self-driving at all, maintaining that an autonomous system must relieve the driver of making decisions in either the longitudinal or latitudinal dimension. Not to get all semantic, but here we quibble. Porsche’s system — which we experienced in a hacked first-generation Panamera Turbo — seemingly controls travel in forward motion, and quite effectively and efficiently we might add.

The system takes Porsche's adaptive cruise control — similar to others, which follow at or below a set speed, maintaining a safe distance based on the radar-measured velocity and behavior of the car being followed — and builds on it. It adds a wealth of satellite-derived geographic and road congestion information, as well as cues about posted speed limits and road condition garnered from its cameras, radar and a host of onboard sensors.

What this translates to is a car — a hypothetical, futuristic car — that can predict what is going to happen on the roadways and adjust throttle, brake, and transmission outputs to optimize performance and/or efficiency. So, if a sweeping right hand turn is coming up, followed by an uphill stretch, and capping off by a reduction in speed limit at the crest, the InnoDrive-equipped vehicle will use all of the known data to create a “prediction horizon” regarding optimal outcomes. This allows it to know at precisely what speed to take the corner, when to accelerate on the incline, and when to coast, brake, or drop down a gear and engine brake once at the top. In a pilot experiment, the result was a computer-driven Porsche that was 10% more efficient and 2% quicker than human drivers on the same route.

Once the system is engaged, you keep your feet off the gas and brake pedals. This being a Porsche, you do all of the steering yourself. You also get the choice of selecting between efficient, comfort, and dynamic modes, which grant primacy, respectively, to maximizing fuel economy; minimizing disruptive g-forces, and maximizing velocity of travel (within reason).

Sometimes, all three of these things occur simultaneously, to great effect. As one of the engineers in charge of designing the system told us, “Driving around a corner, you do not gain efficiency by slowing down. So the system will always go around the corner as fast as possible.” Of course, it will do so within the context of the lateral or longitudinal g-force limits embedded in the settings.

On the roads around Weissach, the system worked seamlessly, if a bit surprisingly. Since it was far more “familiar” with the roads—through access to mapping and data sensing — than we were, it was likely to speed up through corners and over blind crests through which we might have initially braked, or to slow down in areas through which we may have been inclined to blast. Fortunately, it was kind enough to present us with graphics explicating its behavior. These little icons, presented in an auxiliary gauge, indicated obstacles or occasions like hairpin turns, or changes in the posted speed limit. Though informative, they were somewhat akin to one’s spouse “asking” for permission to do something that they’re planning to do regardless.

The system may find its highest state when equipped in Porsche’s ever-increasing fleet of hybrid vehicles, where it could boost range or save electric energy. If, for example, we were passing through a city like London that places a heavy toll on cars that emit CO2, the system will save enough charge so that it can make it to our destination without a ticket. More intriguingly, InnoDrive could sport an ultra-dynamic mode to demonstrate to drivers (and passengers) the limits of a car’s capabilities.

Like its ostensible driver, for now InnoDrive is crippled by its inability to see unpredictable objects further out than its 200-meter range of vision. it is also unable (in current prototype form) to tell between red and green traffic lights. “For now,” the engineers warned us, “the system assumes that all lights are green.” This means turns, stoplights, and roundabouts present something of a challenge, as do dozens of other situations for which the code has yet to be written.

We never crashed, but there were times when we felt as though we might. The Porsche InnoDrive system will be most effective when car-to-car communications become possible and widespread — and by that, we mean the kind of car-to-car communication that does not involve impact.