News

New preprint from our autonomy research area!

TreeIRL is a novel planner that combines classical search with learning-based methods to achieve state-of-the-art performance in simulation and in real-world autonomous driving! 🚘 🤖 🚀

💡 The key idea is to use Monte Carlo tree search (MCTS) to find a promising set of safe candidate trajectories and inverse reinforcement learning to choose the most human-like trajectory among them.

Why this matters:

🛣️ First real-world evaluation of MCTS-based planner on public roads.

📊 Comprehensive comparison across simulation and 500+ miles of urban driving in Las Vegas.

🏆 Outperforms classical + SOTA planners, balancing safety, progress, comfort, and human-likeness.

🧩 Flexible framework that can be extended with imitation learning and reinforcement learning.

‼️ Underscores importance of diverse metrics and real-world evaluation.

–> read more about it here.

#AI #Robotics #MachineLearning #SelfDrivingCars #AutonomousVehicles #MotionPlanning

🐭🍔⚡🧠 Our paper studying the neural substrates of patch foraging decisions – whether to stay on a patch or leave it for a more promising one – was accepted to Neuron!

🧠 Our paper studying how the brain reuses strategies from previous tasks to solve novel tasks was published in PLOS Biology. This follows up on our previous work on multi-task reinforcement learning in humans, where we first reported behavioral evidence for such policy reuse.

🚗 Excited to share that our Lab2Car paper was accepted to ICRA 2025! See you in Atlatna!