Google has been actively engaged in green energy for the last several years, but it has been a mixed bag.

The search giant has invested in companies like Alta Rock and Makani Power and has gone to Washington to plead the case of alternative energy. It has obtained a license from FERC to buy and sell electricity to improve its own operations and has championed the cause of energy-efficient power supplies and components in computers.

Some of the company's efforts, however, have been downright wacky. For example, Google retrofitted a series of hybrid cars with extra batteries to show the potential for plug-in hybrids. The cars got around 60 miles per gallon, which is more than the usual Prius but not high enough to justify a retrofit on an economic basis. Suddenly, one of the cars started getting around 100 miles a gallon. The secret? Google drove the car on a closed circuit course with a professional driver. The economics probably don't work on that one, either.

The company has also been on a quest to develop new mirrors for solar thermal power plants, despite the fact that developing these mirrors revolves around chemical engineering and material science. Those are going to be some lonely scientists in the company cafeteria.

The story in The New York Times about the company's robotic car, though, shows how the company wants to apply its core competence -- artificial intelligence -- in the realm of efficiency. The car effectively runs on a robotic navigation and control system powered by the same sort of Bayesian probability algorithms behind its search engine. (The nav system was devised by Stanford professor Sebastian Thurn, who won the DARPA Challenge a few years back. That is not Google's car in the picture. It is Thurn's car Stanley that won the DARPA challenge with financial and other support from MDV and Google, among others.) Fifteen years ago, probability was scoffed at as a basis for artificial intelligence, but now it practically dominates the field. Google has been one of the more visible champions of the concept.

Nissan, Toyota, Ford and others have worked on drive-by-wire systems for years as a way to reduce traffic accidents and improve mileage. Here's a video of Nissan's Eporo robots that space themselves out to prevent traffic bulges and collisions.

Overall, the navigation system fits in much better with Google's strengths than VC investing or solar thermal. Presumably, you can expect more of this sort of thing. PowerMeter, for instance, could be enhanced to become a home automation system that feeds data about your personal habits (e.g., homeowner opens fridge at 7 a.m., 12:15 and 6:25 p.m.; dries clothes at 8:00 p.m., but doesn't open dryer door until 7 a.m.) into the control system to save energy. PC and commercial building automation could emerge from the same vein. It might even team up with companies doing Bayesian hardware like Lyric Semiconductor to improve the efficiency of cloud systems.

One can even imagine software control systems for reflectors at solar thermal plants. This sort of automated intelligence is at the core of what eSolar does.

So, although robot cars may sound a bit daft at first, the company appears to be aiming closer to home.