Composite shot of testing. (via MIT)
The days of you being in control of your vehicle while encountering road hazards could be coming to an end thanks to some clever researchers over at MIT. In a recent paper, co-authored by PHD student Sterling Anderson and research scientist Karl Iagnemma, the team describes how using their ‘semi-autonomous co-pilot system’ can take control of your vehicle for obstacle avoidance.
The autonomous avoidance system the team designed uses an on-board camera system in conjunction with a laser range-finder that is able to identify road hazards situated in the vehicles environment. The researchers devised software, based off of homotopy, which takes the information provided by the camera and range-finder and plots a safe course around the identified hazards. Homotopy is a form of topology which takes two continuous functions from one topical space to another that can be ‘continuously deformed into one-another’. This means that the software looks for the best path based on the vehicle's direction, speed and hazard and plots a rout that diverges from the current location around the obstacle to an appointed/identified safe-zone and then back to the original path or direction.
During the hazard diversion, the software takes control of a series of actuators that are connected to the vehicle's steering wheel and its braking system to ‘adjust’ the direction and speed of the vehicle while avoiding obstacles. The team has run over 1,200 test runs of the system using everything from an RC buggy to a Jaguar Type S with promising results. Only a few collisions were due to the camera failing to identify certain hazards. Actually, the system is capable of full autonomous control of the vehicle but the researchers state that the ‘beauty’ of the system is the seamless shared control between the robotic co-pilot and its human counter-part, which is more difficult to achieve over full autonomy.
The system could be advantageous for those driving in severe weather as well as for those that fall asleep while driving or under the influence of alcohol or drugs. (Not giving an excuse to drive in those conditions.) The co-pilot system is still being developed, but the team hopes to have it commercially available in the next 5 years at a cost of $5,000 to $10,000 US. They are also looking at the possibility of using smartphones instead of the systems embedded technology so that anyone could implement the co-pilot into any vehicle. This would mean using the phones cameras, gyroscope and accelerometer as a replacement of the co-pilots avoidance technology; however this too is also a few years away from becoming commercially available.
Cabe
