Tesla in Autopilot Mode Crashes into Another Car Again

No major casualty in Crash

A Tesla vehicle alleged to have been on autopilot ran into a parked police vehicle in California a few days ago. The crash left the Tesla’s driver with minor injuries. Fortunately, the Police car had no one in it at the moment of the crash. Laguna Beach Police department’s public information officer, Jim Cota, however, reports that both vehicles are in bad shape.

He mentioned that the front of the Tesla was badly beat up while the police car was a total loss with axle damage unworthy of repair.  The man behind the wheel of the Tesla told the officers that the car was running in a semi-autonomous mode, but further investigations have to be made to ground this claim.

Tesla vehicle running on autopilot crashed into a stationary Police car

Accidents involving similar autonomous vehicles

If the driver’s claim is confirmed, the Laguna Beach accident will become another number in the statistics of Teslas on autopilot that have crashed into parked emergency vehicles since the year began. It will be recalled that a Tesla collided with a stationary firetruck on the 404 freeway in L.A. county in January and a similar model also crashed into a stationary firetruck in Utah sometimes earlier this year.

Tesla Model S collides with a stopped firetruck in Los Angeles

The automobile company has emphatically maintained that the accidents are majorly caused by human errors and not the autopilot feature. The company emphasizes that drivers have been constantly informed that the autopilot system is in place to assist and complement the efforts of human drivers and not take the place of alert human drivers.  Speaking on behalf of the automobile company, a spokesman mentioned that when making use of the autopilot feature, drivers have constantly been reminded that they should never take their hands off the wheel.

He also added this as the reason why drivers are expected to give an express accept markings, a dialogue box that clearly states that autopilot is designed only for use on highways that have clear lane markings and a certain divider before they make use of the autopilot feature. He further chipped it in that the fact that Tesla is said to be autopilot does not mean the cars are immune to accidents.

It was however reported in April that safety advocates were becoming highly skeptical of Tesla’s claims that autopilot mode reduced crashes by 40%. They regarded this fact as a misrepresentation of the Highway Traffic Safety Administration’s data.

Autopilot systems have common flaws

Wired explains that a couple of autonomous and semi-autonomous systems have a common shortcoming which may be pegged as the cause of the various crashes. The analysis states that the major reason underlying this shortcoming is that the systems are designed to ignore static obstacles.  This is because the systems are usually built in such a way that they avoid false positives and slamming on the brakes unnecessarily. However, slamming on the brakes unnecessarily can be as dangerous as not slamming brakes when necessary.

This is the probable reason for Tesla’s manual express warning, that the available Cruise Control for traffic awareness is unable to identify all objects and may therefore not brake for vehicles or objects which are stationary, particularly in situations where the driver is navigating over 80km per hour and if a vehicle being followed by the Tesla detours from its original path and another stationary car  or object is ahead of the car.

It is a similar situation when Volvo’s  Pilot Assist system which is also semi-autonomous gets activated, such that if a car in front of a Volvo using pilot assists switches lane in bringing a stopped vehicle ahead of the car, pilot assist will ignore the parked vehicle and accelerate to the already programmed cruise control speed.

According to Raj Rajkmur, a researcher on the dynamics of autonomous driving at Carnegie Mellon University, the radars used by the system are same as those used in adaptive cruise control systems and are designed to detect only moving objects and may be unable to detect stationary objects.

The best way to tackle this shortcoming is to load the system program with diverse features that enables it to concentrate on moving objects and ensure that the car does not stop every time it comes in contact with a traffic sign. In addition, manufacturers of automobiles can also improve system accuracy by utilizing multiple sensors types such as lidar. Lidar has the ability to project lasers from which objects bounce off to create a  map of the environment which has high resolution and even though they are more expensive than cameras and radar, they are fully capable of sustaining visibility where other sensors fail.