Tata Institute of Fundamental Research
Online Parameter Estimation for Human Driver behavior Prediction from Driving Demonstration Data
STCS Seminar
Speaker:
Raunak P. Bhattacharyya (Stanford University, USA)
Organiser:
Himanshu Asnani
Date:
Thursday, 2 Jan 2020, 14:30 to 15:30
Venue:
A-201 (STCS Seminar Room)
(Scan to add to calendar)
Abstract:
Abstract: Driver models are invaluable for safety validation in simulation. However, driver modeling is characterized by a high degree of uncertainty. While rule-based driver models have the advantage of being interpretable and collision-free, black-box models are more expressive than rule-based models and capture more nuanced behavior. Unfortunately, such black-box models lack interpretablity and fail to incorporate the safety guarantees of the rule-based models. While most approaches in driver-modeling literature select model parameters offline, online estimation has the advantage of being able to capture the behavior of individual drivers. In this paper, we show that online parameter estimation applied to the Intelligent Driver Model captures nuanced individual driving behavior while providing collision free trajectories. We benchmark performance against rule-based and black-box driver models on two real world driving data sets. We evaluate the closeness of our driver model to ground truth data demonstration and also assess the safety of the resulting emergent driving behavior.
| Speaker: | Raunak P. Bhattacharyya (Stanford University, USA) |
| Organiser: | Himanshu Asnani |
| Date: | Thursday, 2 Jan 2020, 14:30 to 15:30 |
| Venue: | A-201 (STCS Seminar Room) |
(Scan to add to calendar)
Abstract:
Abstract: Driver models are invaluable for safety validation in simulation. However, driver modeling is characterized by a high degree of uncertainty. While rule-based driver models have the advantage of being interpretable and collision-free, black-box models are more expressive than rule-based models and capture more nuanced behavior. Unfortunately, such black-box models lack interpretablity and fail to incorporate the safety guarantees of the rule-based models. While most approaches in driver-modeling literature select model parameters offline, online estimation has the advantage of being able to capture the behavior of individual drivers. In this paper, we show that online parameter estimation applied to the Intelligent Driver Model captures nuanced individual driving behavior while providing collision free trajectories. We benchmark performance against rule-based and black-box driver models on two real world driving data sets. We evaluate the closeness of our driver model to ground truth data demonstration and also assess the safety of the resulting emergent driving behavior.