3Qs: The driverless car

May 14, 2012

Marty Vona, an assistant professor in the College of Computer and Information Science, is developing robots that can detect uncertainty in their environment, a particular concern for driverless cars on a road full of unpredictable people.

Nevada recently issued Google the first-ever offi­cial license to operate its self-driven car. Cur­rent laws in other states tech­ni­cally pre­vent cars from dri­ving themselves. We asked Marty Vona, an assis­tant pro­fessor in Northeastern’s Col­lege of Com­puter and Infor­ma­tion Sci­ence, to explain the tech­nical chal­lenges of dri­ver­less cars and how these vehi­cles may change the traffic landscape.

From a robotics perspective, how does the self-driving car work?

To my knowl­edge, Google has released few tech­nical details of its imple­men­ta­tion — its project was a com­pany secret until fall of 2010. Google’s approach is said to com­bine sensor inputs from cam­eras, radar, wheel rota­tion and lasers that have been added to the car, along with pre­vi­ously stored maps and images from the Google Street View data­base. Pre­sum­ably Google also uses GPS to get a rough idea of the car’s loca­tion. Exactly how it is using that infor­ma­tion has not been dis­closed, but it is likely both a com­bi­na­tion of high-level plan­ning and lower-level control.

For plan­ning, the car is prob­ably given a des­ti­na­tion and then plans a route based on its cur­rent loca­tion and map data. The plan could also be revised along the way, should road clo­sures or heavy traffic pro­vide impediments.

For con­trol, which is likely the more chal­lenging task, it would be nec­es­sary to rapidly process data from the onboard sen­sors both to mea­sure nom­inal aspects of the envi­ron­ment, such as road mark­ings, and off-nominal events, such as unex­pected pedestrians. The soft­ware would then need to quickly decide how to modify the car’s dri­ving inputs (steering, throttle, brakes) to react safely.

The legal and eth­ical ques­tions of dri­ver­less cars are very sig­nif­i­cant: Who will be respon­sible —and liable — if a dri­ver­less car is involved in an accident?

Though Google’s car has been making head­lines since it was revealed, self-driving cars have been studied world­wide since the 1980s; some projects had already com­pleted thou­sands of self-driven miles as early as 1995. The pace of advance­ment picked up over the last decade, in part because the United States Defense Advanced Research Projects Agency spon­sored sev­eral high-profile competitions.

What technical challenges does Google face in designing its driverless car?

The main chal­lenges are sen­sors that can sub­sti­tute for human eyes and ears and soft­ware that can reli­ably process that data to drive safely.

The soft­ware would nom­i­nally need to under­stand the cur­rent loca­tion of the car and fun­da­mental aspects of the dri­ving envi­ron­ment, including road bound­aries as well as the posi­tions and speeds of nearby cars and pedestrians. It would also need to under­stand how the car would react to changes in the dri­ving inputs. How far, for example, would the car travel before coming to a stop when a cer­tain level of braking is applied?

A fun­da­mental and deep chal­lenge would be to pre­dict the behavior of sur­rounding vehi­cles, espe­cially when humans may be dri­ving, but there are also a number of more subtle off-nominal sit­u­a­tions that are likely huge chal­lenges to handle auto­mat­i­cally. These chal­lenges include road damage, weather con­di­tions, nearby acci­dents, unusual sig­nage and sudden engine, wheel and break malfunctions

What are the benefits and drawbacks to introducing a self-driving car into the current traffic pattern?

One of the ben­e­fits that has been sug­gested is the pos­si­bility that self-driving cars could reduce acci­dents, many of which are attrib­uted to human error or inattentiveness. Another ben­efit could be reduced traffic delays, since self-driving cars could poten­tially coor­di­nate road use by acquiring and using larger-scale infor­ma­tion about other cars on the road via wire­less com­mu­ni­ca­tion. This would likely require that at least a sig­nif­i­cant frac­tion of all cars be self-driven and par­tic­i­pate in such coordination.

The main draw­backs are safety, lia­bility and cost. Can the cur­rent soft­ware respond as well — and as rapidly — to off-nominal sit­u­a­tions as an expe­ri­enced human? When a self-driven car is involved in an acci­dent, could the maker of the dri­ving system be liable, just as a human driver could be? How much can the cost for the req­ui­site advanced sen­sors and com­puters, which is cur­rently in the tens of thou­sands of dollars,be reduced?

Photo by Mary Knox Merrill