Indian autonomous driving company Swaayatt Robots has announced that it achieved the world’s first Level 5 autonomous driving capability. In the demonstration, its autonomous vehicle i.e, Mahindra Bolero, learned to negotiate complex traffic dynamics in the Toll-Plaza and successfully crossed highly unstructured toll-gates.
The company said that this achievement, second only to previous demos in 2017 and 2023, signifies a significant leap in Level-5 capabilities. The team enabled autonomous driving to negotiate tight and dynamic adversarial environments using multi-reinforcement learning agents. Notably, the demo on October 22, 2023, involved bidirectional traffic negotiation on single-lane roads.
Swayaatt Robots posted a video which displays the vehicle entering the toll-gate region through a highway, navigating bidirectional traffic dynamics in an open area without strict driving rules. The complexity is heightened by randomly parked trucks, a common occurrence in Toll-Plaza areas, and the vehicle’s ability to decide which toll-gate-passage to commit to while avoiding overtaking trucks.
To illustrate the complexity, a tractor tire was placed behind a large 18-wheeler truck, testing the vehicle’s capability to detect obstacles at night. Impressively, the vehicle successfully negotiated this scenario while another truck overtook from the right.
The vehicle demonstrated advanced decision-making and motion planning algorithms, crucial for navigating such intricate scenarios. The team emphasized the scalability of this framework with unsupervised deep learning and teased an upcoming showcase in February, promising an end-to-end negotiation of daytime traffic.
Notably, the vehicle slowed and paused approaching a speed-breaker, aligning itself on-the-fly with the selected toll-gate, adhering to driving rules and speed limits. The system even adapted to unexpected obstacles, such as a broken traffic-police barricade, showcasing a robust and adaptive autonomous driving system.
Interestingly, Tesla cars currently fall under Level 2 of the six levels of vehicle automation defined by the Society of Automotive Engineers (SAE).