The ParkShuttle Collision
On December 6, 2005 two empty ParkShuttle vehicles collided near a one-lane bridge on the Rivium system outside of Rotterdam. The following statement was posted on the transport-innovators discussion list by Robbert Lohmann of Frog Navigation Systems, the ParkShuttle supplier:
During morning operations one of the vehicles experienced a communication failure. As part of the fail-safe protocol this vehicle came to a stop. Because of the location it stopped at, several zones within the system remained blocked, preventing access by other vehicles.
At that time the local crew went out to the vehicle with the communication failure to direct it to a different location. This was done according to procedure. To release the zone a manual action had to be taken by the operator. In this case the operator entered a wrong command - which in first instance did not have any consequences. As the same (wrong) command was re-entered several seconds later, the situation where the collision could occur was created.
All vehicles are equipped with several short range (ultra-sonic) and two long range (laser) obstacle detection systems. The systems on both vehicles reacted properly and did initiate the emergency stop. However, as the line of sight for the sensors at this specific location is limited, the detection is programmed for static objects - knowing that a dynamic object in the same zone is ruled out by the zone administrating system. Unfortunately, the wrongly manual command introduced a dynamic obstacle - the braking distance to come to a complete stop was too long, thus leading to the collision.
Every cloud has a silver lining: the bumpers worked as designed and absorbed, together with the front part of the chassis and body, all the energy of the collision. The passenger compartment is still completely intact. The damage to the vehicles is actually minor (in contrast to what appears on the picture) and can/will be fixed quickly.
Currently we are developing additional measures to prevent such events from (re-)occuring. A consistency check is developed to check continuesly if there is only one vehicle in each zone. Additionally measures will be taken in the software to recognize wrongly manual inputs, not execute this command and warn the operator of the fact that he/she has entered a wrong command.
We are expecting to start testing the additional measures taken in the beginning of January. The operator, Connexxion, will determine when passenger service resumes. Our best guess at this time is that this could be any time between mid-January and end-January.