The Taipei City Government and Turing Drive Inc. have teamed up to introduce a new autonomous driving shuttle service along Xinyi Road during off-peak hours, starting from May 2020. The collaboration began with the signing of an MOU in 2019, and Turing Drive proposed a 3-phase plan for the project.
Phase 1 involved preparing the necessary infrastructure by April, which included creating detailed HD maps, installing roadside units to enable communication between vehicles and infrastructure, setting up electric charging stations, and installing signboards.
Phase 2 focused on testing the operation, which started in May. The autonomous vehicle (AV) system was trained to get comfortable with different driving situations during this phase.
Phase 3, scheduled to begin in September, aimed to launch the actual passenger services if everything went smoothly during the testing phase.
The shuttle service would run back-and-forth on dedicated bus lanes along Xinyi Road, covering a distance of 12.3 kilometers from Zhongshan S. Rd. intersection to Keelung Rd. intersection. At the turning points on open roads, safety drivers would take control of the shuttles.
Turing Drive is a Taiwan-based startup specializing in developing autonomous driving systems. They have formed alliances with Tron Energy Corporation, Trillion, ThinkTron, International Integrated Systems (IISI), and Metropolitan Transport Corporation. Interestingly, Turing Drive and its team are responsible for producing 100% of the software and 70% of the hardware in Taiwan, which is a remarkable achievement among local companies.
Three electric autonomous Vehicles
The shuttle service would initially use three electric autonomous vehicles, including a 4-meter shuttle with a capacity of 9 passengers and a 6-meter shuttle that can accommodate up to 34 passengers. If certain conditions were met, the maximum speed and operating hours could be increased later on.
Through this joint initiative, the Taipei City Government and Turing Drive aimed to validate autonomous driving technologies and explore the potential of autonomous shuttles as a new public transport solution. The project was expected to stimulate Taiwan’s autonomous driving industry and provide a safe and convenient alternative for transportation during off-peak hours.
Is a self-driving bus in a smart city just a distant futuristic dream? Or are the concerns about connected cars and their potential dangers something that only Tesla drivers and researchers need to worry about in the far-off future?
In downtown Taipei, Taiwan, a self-driving bus is currently undergoing testing. It first hit the road in May, but without passengers. Now, it has progressed to the next phase of its trial run, from October to February 2021, where it’s carrying passengers, citizens, and even foreign visitors.
This electric bus, developed and operated by Turing Drive, comes equipped with advanced technology, including an HD map, GNSS (Global Navigation Satellite System) receivers, lidar (light detection and ranging) sensors, cameras, and radars. To ensure safety, the bus receives real-time traffic light status information from InVignal and uses two V2X (vehicle-to-everything) roadside units to detect potential collisions.
The bus follows a 12.3-km route on a dedicated bus lane daily from 12:30 a.m. to 2:30 a.m. (0:30 to 2:30). Even during off-peak hours, it was in high demand, and I had to wait for seven days to get a free trial-ride.
When my turn finally came, I joined other passengers at the assigned stop. I noticed how the bus smoothly came to a halt.
Before boarding, the observer scanned my QR code. This manual process will be replaced with an automated identification system when the bus becomes fully operational, and the observer role will be eliminated once the trial runs are complete.
Once inside, I saw a large screen displaying real-time parameters collected by the sensors, as well as a high-definition map with white lines representing the route.
CAN (Controller Area Network)
The bus operates on a Linux system, controlled via a CAN (Controller Area Network) bus. The GPU (graphics processing unit) calculates acceleration and braking decisions. The system worked well, although the driver had to manually intervene near traffic lights because the communication system displayed all red lights ahead. Once past the lights, the bus resumed autopilot mode. The driver joked that we were lucky, as the system had been functioning normally for a long time before our ride.
Overall, the bus accelerated and braked smoothly while following the mostly linear bus lanes. However, the road bumps were more noticeable due to the slow speed. As it was midnight, there were no other vehicles or pedestrians to thoroughly test the automatic braking system.
Despite the speed limit, the bus reached the next stop safely in around 15 minutes. All passengers received a souvenir bus ticket. My experience was mostly smooth, aside from the traffic light hiccup. Research revealed that some test runs experienced sudden braking when the bus detected cars in adjacent lanes, but Turing Drive promptly addressed the issue the next day.
During regular days and busy hours when traffic is heavy and pedestrians are crossing the streets. I eagerly look forward to my next ride and the day when this self-driving bus becomes a common sight on the roads during the daytime.
Below are some photos I took during the ride, showcasing the bus’s mechanisms and my souvenir ticket.
The idea of self-driving buses navigating roads and safely transporting passengers from one location to another is truly remarkable.
Research into defense solutions for connected cars has highlighted the fact that this is a relatively new and unexplored realm, making the potential dangers largely speculative. Despite being in the testing phase, the presence of these buses on the road could bring anticipated threats closer to reality.
Threats to connected vehicles revolve around the mechanisms enabling them to drive autonomously and gather data from their surroundings. One notable concern is the CAN bus component, which could transmit malicious messages if exploited by threat actors. Attackers might abuse default system configurations, jam radio transmissions, or execute man-in-the-middle (MitM) wireless data transmissions.
Manufacturers and developers can invest efforts
In integrating security measures into the connected technologies that will drive various types of smart vehicles, including self-driving buses, in the future. In the meantime, implementing general security recommendations can be beneficial:
Implement an IDS/IPS for the CAN bus: These are network security systems that analyze traffic flow to detect and prevent network attacks. An IDS/IPS for the CAN bus monitors the vehicle’s network for suspicious messages using deep packet inspection.
Strengthen defenses against lidar and radar attacks: Research has shown potential vulnerabilities in lidar and radar sensors that could be exploited. Ensuring self-driving models are robust enough to fend off such attacks is crucial.
Secure the operating system: Given that the bus operates on Linux and CUDA applications, protective measures should be taken, just like any other operating system, to safeguard against threats.