A machine learning approach for personalized autonomous lane change initiation and control

Abstract

This paper focuses on the design of hierarchical control architectures for autonomous systems with energy constraints. We focus on systems where energy storage limitations and slow recharge rates drastically affect the way the autonomous systems are operated. Using examples from space robotics and public transportation, we motivate the need for formally designed learning hierarchical control systems. We propose a learning control architecture which incorporates learning mechanisms at various levels of the control hierarchy to improve performance and resource utilization. The proposed hierarchical control scheme relies on high-level energy-aware task planning and assignment, complemented by a low-level predictive control mechanism responsible for the autonomous execution of tasks, including motion control and energy management. Simulation examples show the benefits and the limitations of the proposed architecture when learning is used to obtain a more energy-efficient task allocation.

A Novel Lane-Changing Decision Model for Autonomous Vehicles Based on Deep Autoencoder Network and XGBoost

“…In addition to the microscopic running states of the subject and the adjacent vehicles, a driver with an LC motivation needs to immediately make the optimal LC decision according to the other related factors [53]. In this study, we selected 17 candidate features based on the experience [20], [29] of researchers in feature selection, as shown in Table 1. The vehicle motion parameters are directly transformed into vectors as the model input, which might result in data redundancy.…”

“…Xu et al [28] proposed a fusion LCD model based on a gradient boosting decision tree (GBDT) and compared the effects of different features on the decision results. The XGBoost algorithm, which is based on the boosting tree, further improves the loss function, regularization and parallelization processes and exhibits excellent classification performance, which can effectively improve the precision of the LCD model [29]- [31].…”

A Reinforcement Learning Based Approach for Automated Lane Change Maneuvers

“…Machine learning (ML) algorithms have the capability of dealing with unforeseen situations after being properly trained on a large set of sample data without explicitly specific design and programming rules beforehand. Vallon et al [7] proposed to use Support Vector Machine to make the lane change decision tailored to the driver's individual driving preferences. After the lane change demand is generated, the maneuver is executed using a MPC.…”

A Novel Trajectory Planning Method for Automated Vehicles Under Parameter Decision Framework

“…In decision-making problems, the common method is rule-based and includes scenario-based state machines [3], [4] and Markov decision process models [5], which are predefined by experts [6], [7]. Otherwise, a learning model trained by driving data is used [8], [9]. The decisions in these studies are specific and predefined driving behaviors in a finite set, such as overtaking, lane-maintaining, lane-changing, right-hand turning, etc.…”