Core Concepts
Efficient path planning under sensing uncertainty is crucial for off-road autonomous vehicles, balancing traversal time and collision cost.
Abstract
The article discusses the challenge of long-range dynamic replanning for off-road autonomous vehicles in uncertain environments. It introduces the DREAMS algorithm, which leverages multi-sample posterior sampling to improve planning efficiency. The key focus is on balancing traversal time and collision cost while considering uncertainty in perception. By evaluating multiple plausible optimal paths and worlds, DREAMS outperforms other determinization-based approaches. The study compares DREAMS with DRPS and Sampled A* algorithms on real-world datasets from the DARPA RACER program, demonstrating superior performance in terms of combined traversal time and collision cost. The research highlights the importance of considering uncertainty in long-range path planning for off-road autonomous driving.
Stats
"Our method, Dynamic Replanning via Evaluating and Aggregating Multiple Samples (DREAMS), outperforms other determinization-based approaches in terms of combined traversal time and collision cost."
"We evaluate algorithms for dynamic replanning on a large real-world dataset of challenging long-range planning problems from the DARPA RACER program."
"With this framework, DREAMS enables reasoning not just over the distribution of worlds but also over additional parameters such as traversal speed."
"DREAMS plans effectively under uncertainty to achieve lower total cost compared to other determinization methods."
"On a large dataset of challenging long-range planning problems, we demonstrate that DREAMS plans effectively under uncertainty to achieve lower total cost compared to other determinization methods."
Quotes
"Our key insight is that this deficiency stems from determinization’s limited ability to reason about the distribution of costs over plausible worlds."
"With this framework, DREAMS enables reasoning not just over the distribution of worlds but also over additional parameters such as traversal speed."
"DREAMS plans effectively under uncertainty to achieve lower total cost compared to other determinization methods."