This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
An Improved, Autonomous, Multimodal Estimation Algorithm to Estimate Intent of Other Agents on the Road to Identify Most Important Object for Advanced Driver Assistance Systems Applications Using Model-Based Design Methodology
- Mayur Rajendra Waghchoure - Dorle Controls Pvt Ltd, Driver Assisted Systems, India ,
- Kunj Pareshkuma Patel - DEKA Research & Development, USA ,
- Vysali Nikhita Rachamadugu - Ford Motor Company, USA ,
- Muhammad Hameem Safwat Husain Javid Iqbal - DEKA Research & Development, USA ,
- Sai Kamal Sreeja Veepuri - Motional, USA ,
- Bhargav Narsinha Deshpande - Dorle Controls Pvt Ltd, USA ,
- Aniruddha Dorle - Dorle Controls LLC, USA
Journal Article
12-06-01-0005
ISSN: 2574-0741, e-ISSN: 2574-075X
Sector:
Citation:
Waghchoure, M., Patel, K., Rachamadugu, V., Iqbal, M. et al., "An Improved, Autonomous, Multimodal Estimation Algorithm to Estimate Intent of Other Agents on the Road to Identify Most Important Object for Advanced Driver Assistance Systems Applications Using Model-Based Design Methodology," SAE Intl. J CAV 6(1):51-82, 2023, https://doi.org/10.4271/12-06-01-0005.
Language:
English
Abstract:
Advanced Driver Assistance Systems (ADAS) are playing a significant role in
enhancing driver safety and occupant comfort in modern vehicles. The primary
research focus in this domain includes the precise perception of the current
state and the prediction of the future states of dynamic agents. To perform
these tasks an intelligent agent capable of operating in the stochastic
environment is implemented in the form of various ADAS features. A trajectory
prediction problem can be defined using either a model-based or data-driven
approach. The current article addresses the problem of trajectory prediction in
the stochastic environment using a model-based approach with a quintic
polynomial as a function approximator to ensure smooth acceleration trajectory
for the left and right lane-change maneuvers. The task of trajectory prediction
also considers the information about the vehicle dynamics, the concept of
Receding Time Horizon (RTH), and the variable curvature model of the road.
Further, the task of assessing the intent of other agents is framed as a Markov
decision process problem due to uncertainty in the agent’s action. The
information about the predicted trajectory and current state of an agent is
processed in the state transition probability estimation module to infer
information about the stochastic policy of other agents in the environment using
a Naïve Bayes classifier algorithm. The decision made by other agents has been
propagated further to the Collision Detection Module to find the Most Important
Object (MIO) for various ADAS features. In this way, the current article
outlines a robust method to predict the intent of non-ego vehicles and identify
the MIO on the road for the ego vehicle to implement various ADAS applications.
The capability of the proposed algorithm to handle both uncertainties in the
environment and in decision-making by other agents has been validated with
several simulated driving scenarios.