Paper Image

Predicting Human Actions to Enable Safe and Efficient Autonomous Driving

Published on:

13 March 2021

Primary Category:

Artificial Intelligence

Paper Authors:

Arjun Sripathy,

Andreea Bobu,

Daniel S. Brown,

Anca D. Dragan

Bullets

Key Details

Proposes algorithm for robot to switch human prediction models online

Estimates performance vs. computation tradeoff for each model

Selects complex models in critical situations, simple ones otherwise

Achieves high rewards with lower computation in simulations

AI generated summary

Predicting Human Actions to Enable Safe and Efficient Autonomous Driving

This paper proposes an approach for autonomous vehicles to dynamically switch between predictive models of human driver behavior. By estimating potential gains in safety and performance, the vehicle selects either a complex, accurate model or a simple, fast one. This allows responding appropriately in critical situations while minimizing computation. The method is tested in simulated driving scenarios, achieving high rewards at low computational cost.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up