Autonomous Driving Off-Roads through Dynamic and Complex Traffic in India
High-speed off-roads autonomous driving, with bidirectional dynamic traffic, paving the road for military and several other off-roads autonomy operations.
The demo and tech-capabilities addresses several of the requirements of not just Indian armed forces, but also addresses the solutions to various problems that are currently a topic of active research for DARPA and US DoD.
In this demo, it can be seen that our autonomous vehicle is able to seamlessly navigate off-roads environment, along with negotiating bidirectional traffic who do not follow the traffic-rules in general.
As per the driving norms a vehicle should drive on the left side. Whenever our vehicle senses that the other impeding dynamic vehicle is following the rule and is navigating on the correct side, it avoids them from the left, and when it senses the other obstacle is coming from the wrong side, rather than getting confused and coming to a halt, it avoids it from the opposite side. This is soft-traffic-rules constraints adherence.
Currently our vehicle can perform this at 44 KM/H, where it slows down for negotiation, coming to safe speeds. This framework is being scaled up further for 60-80 KM/H relative-speed avoidance maneuvers on such roads.
DARPA's RACER program requires #autonomousdriving technology that enables #autonomousvehicles to navigate off-roads at human driven speeds. This demo is a culmination of our previous demos, the one we did in April 2023, showcasing near human-speeds off-roads driving (DARPA's objective), September 2023 off-roads (showcasing generic traffic negotiation), and our generic bi-directional traffic negotiation on single lane roads.
When this R&D project is complete, our vehicle will be able to perform such maneuvers end-to-end killing each and every perception algorithm, purely on the basis of self-acquired skills via abstraction learning, enabled by our novel research directions at Swaayatt Robots (स्वायत्त रोबोट्स) using inverse #reinforcementlearning and semi-MDPs based formulation.
#deeplearning #machinelearning