Tesla’s Full Self-Driving (FSD) system was recently tested for its ability to navigate the complexities of a multi-story carpark at the Commercial Club in Albury, Australia. This location features 16 Tesla Superchargers, making it one of the largest charging stations in the country. The challenge involved not just reaching the carpark but also successfully navigating the steep ramp to the first-floor charging area.
Upon entering the destination of the Albury Supercharger into the navigation system, the vehicle drove to the Commercial Club. However, instead of proceeding into the carpark on the ground floor, it parked on the street adjacent to the venue. This initial move indicated that the FSD was not equipped to complete the journey autonomously from beginning to end.
To continue the test, the driver manually entered the carpark and engaged the FSD feature. At this point, the system attempted to navigate out of the carpark instead of fulfilling its intended task. After re-entering the carpark and positioning the vehicle near the on-ramp, the FSD was activated again. This time, it successfully guided the car up the steep ramp.
As the vehicle approached the top of the ramp, it came to a halt. The FSD indicator displayed a notable change, shrinking from a prominent position to a more subdued one, indicating a lack of confidence from the system. The incline created a blind spot, making it difficult for the vehicle to assess the area ahead. Recognizing this limitation, the human driver pressed the accelerator, allowing the car to gain visibility and proceed to the first level where the Superchargers were located.
Interestingly, rather than directing the vehicle to a Supercharger bay, the system continued to navigate up additional ramps, eventually reaching the top floor of the carpark. Upon arriving at the geographical coordinates of the Supercharger, the navigation destination was cleared, leading the vehicle to enter a free roam mode instead of attempting to park in a designated charging space.
After charging the vehicle, the driver set a new destination for exiting the carpark. The FSD successfully navigated the turns necessary to descend the exit ramp and maneuver through the corners, ultimately leaving the multi-story structure. However, the system encountered difficulty at a junction, misjudging the direction and turning left instead of right. Fortunately, there was no traffic to impede the vehicle’s path, but this miscalculation highlighted an area for improvement.
The decision-making process for the left versus right turn was not immediately clear, even to the human driver, until lane markings became visible after the turn. This experience underscores the need for enhanced clarity in the system’s navigation decisions in similar scenarios.
Overall, the test revealed that while Tesla’s FSD demonstrated competence in detecting other vehicles, pedestrians, and road features, it requires refinement in its confidence levels when faced with limited visibility. Additionally, the navigation system may benefit from incorporating a better understanding of vertical movement, as it currently appears to rely on two-dimensional coordinates without fully accounting for the multi-level nature of carparks.
For those interested in a visual representation of the testing experience, a video showcasing the FSD’s performance is available for viewing.
