Decision making for self‐driving cars is usually tackled by manually encoding rules from drivers’ behaviours or imitating manipulation using supervised learning techniques. Both of them rely on mass driving data to cover all possible scenarios. This study presents a hierarchical reinforcement method decision cars, which does not depend large amount labelled data. comprehensively considers both high‐level manoeuvre selection and low‐level motion control in lateral longitudinal directions. The...

In reinforcement learning (RL), function approximation errors are known to easily lead the Q -value overestimations, thus greatly reducing policy performance. This article presents a distributional soft actor-critic (DSAC) algorithm, which is an off-policy RL method for continuous control setting, improve performance by mitigating overestimations. We first discover in theory that distribution of state-action returns can effectively mitigate overestimations because it capable adaptively...

In a platoon control system, fixed and symmetrical topology is quite rare because of adverse communication environments continuously moving vehicles. This paper presents distributed adaptive sliding mode scheme for more realistic vehicular platooning. this scheme, mechanism adopted to handle parametric uncertainties, while structural decomposition method deals with the coupling interaction topology. A numerical algorithm based on linear matrix inequality developed place poles motion dynamics...

This paper proposes a longitudinal platoon controller for connected vehicles (CVs) by considering the information of multiple preceding and car-following interactions between CVs. The stability proposed is analyzed using Routh criterion. For verification, we develop an integrated control framework CVs in V2V/V2I communication environment. consists two main components: simulation platform experimental platform. In particular, developed based on TransModeler software, designed self-developed...

Abstract Recent years have seen the rapid development of autonomous driving systems, which are typically designed in a hierarchical architecture or an end-to-end architecture. The is always complicated and hard to design, while more promising due its simple structure. This paper puts forward method through deep reinforcement learning algorithm Dueling Double Deep Q-Network, making it possible for vehicle learn by itself. firstly proposes lane-keeping task. Unlike traditional image-only state...

Decision and control are core functionalities of high-level automated vehicles. Current mainstream methods, such as functional decomposition end-to-end reinforcement learning (RL), suffer high time complexity or poor interpretability adaptability on real-world autonomous driving tasks. In this article, we present an interpretable computationally efficient framework called integrated decision (IDC) for vehicles, which decomposes the task into static path planning dynamic optimal tracking that...

The uncertainties arising from the plant model and topologies have been a major challenge in multiagent consensus control. This article presents distributed robust control method for an uncertain system with eigenvalue-bounded topologies. heterogeneity of node dynamics is described as linear common certain part. transformation adopted to decompose topologically coupled controllers. Then, matrix inequalities (LMIs) technique used numerically solve controller problem. It proved that such...

Dynamic/Kinematic model is of great significance in decision and control intelligent vehicles. However, due to the singularity dynamic models at low speed, kinematic have been only choice under such driving scenarios. Inspired by concept backward Euler method, this paper presents a discrete bicycle feasible any speed. We further give sufficient condition, based on which numerical stability proved. Simulation verifies that (1) proposed numerically stable while forward-Euler discretized...

Connected vehicles will change the modes of future transportation management and organization, especially at an intersection without traffic light. Centralized coordination methods globally coordinate approaching from all sections by considering their states altogether. However, they need substantial computation resources since own a centralized controller to optimize trajectories for in real-time. In this paper, we propose scheme automated signals using reinforcement learning (RL) address...

Abstract Management of connected vehicles at unsignalised intersections is a large‐scale complex problem with safety constraints and time‐varying unsolved variables, which crucial but hard to solve online. A faster coordination system, however, not only benefits from smaller time granularity find optimum, also has more robustness towards scenario fast‐moving vehicle nodes. This paper proposes real‐time scheme consisting three stages. (a) Target velocity optimisation: collision‐free passage...

Reinforcement learning (RL) has achieved remarkable performance in numerous sequential decision making and control tasks. However, a common problem is that learned nearly optimal policy always overfits to the training environment may not be extended situations never encountered during training. For practical applications, randomness of usually leads some devastating events, which should focus safety-critical systems such as autonomous driving. In this paper, we introduce minimax formulation...

Reinforcement learning (RL) algorithms have been successfully applied to a range of challenging sequential decision-making and control tasks. In this paper, we classify RL into direct indirect according how they seek the optimal policy Markov decision process problem. The former solves by directly maximizing an objective function using gradient descent methods, in which is usually expectation accumulative future rewards. latter indirectly finds solving Bellman equation, sufficient necessary...

The conflict between limited road resources and rapid car ownership makes the traffic signal timing become a pivotal challenge. Emerging studies have been carried out on adaptive timing, but most of them still focus throughput intersections, leaving safety travel experience unconsidered. This paper proposes time difference penalized method by reinforcement learning technique to balance capacity in control system. Firstly, microcosmic state representation is proposed integrate dynamics both...

In this paper, a decentralized cooperative adaptive cruise control algorithm for vehicles in the vicinity of intersections (CACC-VI) is proposed. This designed to make use road capacity let more get through intersection within limited traffic signal period consideration safety, fuel consumption, speed limit and heterogeneous features vehicles, passenger comfort. Within platoon, try find optimal input by distributed PSO order reduce tracking errors while respecting different constraints. A...

Traffic congestion is an overwhelming problem faced by road travelers all over the world. A time-efficient and accurate prediction of upcoming traffic can reduce this through enabling proactive planning routes. Recent research suggests that accuracy requires extraction hidden features network from historical data. In general, data either limited (with a longer sampling time) or not provided providers. urban areas, lights, weather conditions, city events, accidents, people's habits...

Dynamic/kinematic model is of great significance in decision and control intelligent vehicles. However, due to the singularity dynamic models at low speed, kinematic have been only choice under many driving scenarios. This paper presents a discrete bicycle feasible any speed utilizing concept backward Euler method. We further give sufficient condition, based on which numerical stability proved. Simulation verifies that (1) proposed numerically stable while forward-Euler discretized diverges;...

To study the influence of eccentricity on dynamic characteristics grinding electric spindle, firstly, nonlinear oil film force (NOFF) expression sliding bearing is established based boundary condition adhesion pressure. Then, unbalanced magnetic pull (UMP) under air gap derived. Finally, bearing-rotor dynamics equation spindle considering NOFF and UMP established, dimensionless processing carried out. The system geometric mass out through shaft center trajectory vibration spectrum. research...