汽车 > 无人驾驶博客



Navya: Where's the Remote?

by Roger Lanctot | 7月 12, 2019

When a commercial truck struck a Navya autonomous shuttle in Las Vegas on November 8, 2017, it set off a torrent of headlines around the world. Now, the National Transportation Safety Board (NTSB) has released its account of the "crash" which resulted in minor damage to both vehicles and no human injuries but revealed important insight regarding local and remote vehicle control.

The Navya autonomous shuttle operated on a predetermined and carefully mapped route and at low speed. The truck driver was backing up at the time of the incident and essentially backed into the Navya, which was stopped.

As an autonomous vehicle, the Navya lacked a steering wheel, brake pedal or accelerator pedal. There was a safety driver on board the vehicle along with a handheld controller - but the handheld controller was not retrieved during the incident.

Since the incident, the operator of the shuttle has made access to the controller easier, according to the NTSB report. It is clear from the report, though, that nothing less than putting the Navya in reverse would have avoided the impact with the truck.

The operation of the vehicle was monitored remotely from Lyon, France, but there was no provision for remote control of the vehicle. It was not clear from the NTSB report whether communication with the safety driver was possible or whether any took place.

There are a couple of important takeaways from what might otherwise be considered a trivial incident:

  1. The meat-head behind the wheel. The truck driver told investigators that he thought or he expected the autonomous vehicle to stop sooner thereby avoiding a collision. This speaks to the expectations of human drivers when interacting with automated vehicles. Humans expect the machines to give way and/or to be easily "intimidated." Maybe Navya forgot to program intimidation into its Shuttle. Ironically the truck driver either failed to notice the Navya passengers waving and trying to get his attention prior to the slow motion crash or chose to ignore them or simply misunderstood what they were trying to communicate.
  2. Monitoring is not enough. Providing for remote monitoring of an autonomous vehicle without providing for remote control is nearly pointless. Navya may have corrected this deficiency on some of its vehicles but the NTSB ought to - but did not - recommend a requirement for remote control.
  3. Remote control is not enough. The Navya vehicle was equipped with a significant compliment of sensors enabling the vehicle to avoid collisions on its own. These systems appear to have functioned appropriately - with the possible exception that the ability to reverse to avoid danger was not programmed into the vehicle.
  4. Some crashes can't be avoided. If an autonomous or human operated vehicle or device or a wild animal is determined to collide with an autonomous vehicle then that collision is going to happen regardless of the cleverness of the technology on the AV. Clearly, the driver of the truck was "determined" to hit the Navya autonomous shuttle once he arrived at the conclusion that the shuttle would stop at a sufficient distance - which it did not.

The only remaining question is why the NTSB did not recommend or comment on the efficacy or requirement for remote control - at least as a topic worthy of further research and investigation. It is the need for remote control that transformed this trivial incident from 2017 into a significant turning point in the history of autonomous vehicles.

The full highway accident brief is available online at  https://go.usa.gov/xyNnR and other publicly released information about the NTSB's investigation of the collision is available online at https://go.usa.gov/xyNnS

Previous Post: Hayden AI: The Future of Law Enforcement | Next Post: GM's Not-so-Super Cruise
Leave a comment