A Georgia Institute of Technology research team has discovered a way to make self-driving cars safe when they’re driven in hazardous road conditions or at high speeds.
Up until now we’ve seen driverless cars performing comfortably on roads in good condition, but by using advanced algorithms and onboard computing, together with installed sensor devices, the Georgia Tech team was able to maintain control of a driverless vehicle when roadway adhesion was limited.
So a driverless car would be able to perform in icy or, as the researchers tested, rally-style conditions.
“An autonomous vehicle should be able to handle any condition, not just drive on the highway under normal conditions,” said School of Aerospace Engineering professor and expert on the mathematics behind rally-car racing control, Panagiotis Tsiotras.
“One of our principal goals is to infuse some of the expert techniques of human drivers into the brains of these autonomous vehicles.”
The Georgia Tech researchers used a method called model predictive path integral control (MPPI) to keep their cars at the edge of their limits.
To create their MPPI control algorithm the team combined large amounts of car handling information with data on the dynamics of the vehicle, to calculate the most stable trajectories from the numerous possibilities.
“Aggressive driving in a robotic vehicle – manoeuvring at the edge – is a unique control problem involving a highly complex system,” said School of Aerospace Engineering assistant professor and project leader, Evangelos Theodorou.
“However, by merging statistical physics with control theory, and utilising leading-edge computation, we can create a new perspective, a new framework, for control of autonomous systems.”
The MPPI control algorithm was tested by racing, sliding and jumping one-fifth scale, fully autonomous auto-rally cars at the equivalent of 90 mph.
The cars carried a motherboard with a quad-core processor, a potent GPU and a battery. Each vehicle was also fitted with two forward-facing cameras, an inertial measurement unit and sophisticated wheel-speed sensors.
In order to maintain balance in the hazardous testing conditions the cars had to balance a desire to stay on the track with achieving the desired velocity.
The researchers refer to these two separate desires, which they managed to coordinate, as costs.
“What we’re talking about here is using the MPPI algorithm to achieve relative entropy minimisation, and adjusting costs in the most effective way is a big part of that,” said James Rehg, a professor in the Georgia Tech School of Interactive Computing.
“To achieve the optimal combination of control and performance in an autonomous vehicle is definitely a non-trivial problem.”