Pan Zhao

Motion planning for autonomous vehicle based on radial basis function neural network in unstructured environment.

The autonomous vehicle is an automated system equipped with features like environment perception, decision-making, motion planning, and control and execution technology. Navigating in an unstructured and complex environment is a huge challenge for autonomous vehicles, due to the irregular shape of road, the requirement of real-time planning, and the nonholonomic constraints of vehicle. This paper presents a motion planning method, based on the Radial Basis Function (RBF) neural network, to guide the autonomous vehicle in unstructured environments. The proposed algorithm extracts the drivable region from the perception grid map based on the global path, which is available in the road network. The sample points are randomly selected in the drivable region, and a gradient descent method is used to train the RBF network. The parameters of the motion-planning algorithm are verified through the simulation and experiment. It is observed that the proposed approach produces a flexible, smooth, and safe path that can fit any road shape. The method is implemented on autonomous vehicle and verified against many outdoor scenes; furthermore, a comparison of proposed method with the existing well-known Rapidly-exploring Random Tree (RRT) method is presented. The experimental results show that the proposed method is highly effective in planning the vehicle path and offers better motion quality.

Lane change path planning based on piecewise Bezier curve for autonomous vehicle

Autonomous vehicle is an efficient component in vehicle active system to reduce traffic accidents. The path planning of lane change for autonomous vehicle is an important component in the autonomous vehicle field. In order to generate a path for passing the front vehicle, a new path planning method that is curvature-continuous and satisfies the vehicle nonholonomic constraint is introduced. Firstly, we calculate the safe lane change distance for autonomous vehicle. Secondly, we generate the path based on piecewise quadratic Bezier curve and the maximum curvature of the Bezier curves is calculated to verify whether the autonomous vehicle can follow this path. Then, we present the indicator based on yaw-rate of vehicle to verify the vehicle ride comfort. The experiment results show that the path generated by this method can be successfully executed by autonomous vehicle, the curvature of the path is continuous, meanwhile vehicle nonholonomic and ride comfort indicator constraint condition is met.

Drivers’ Visual Behavior-Guided RRT Motion Planner for Autonomous On-Road Driving

This paper describes a real-time motion planner based on the drivers’ visual behavior-guided rapidly exploring random tree (RRT) approach, which is applicable to on-road driving of autonomous vehicles. The primary novelty is in the use of the guidance of drivers’ visual search behavior in the framework of RRT motion planner. RRT is an incremental sampling-based method that is widely used to solve the robotic motion planning problems. However, RRT is often unreliable in a number of practical applications such as autonomous vehicles used for on-road driving because of the unnatural trajectory, useless sampling, and slow exploration. To address these problems, we present an interesting RRT algorithm that introduces an effective guided sampling strategy based on the drivers’ visual search behavior on road and a continuous-curvature smooth method based on B-spline. The proposed algorithm is implemented on a real autonomous vehicle and verified against several different traffic scenarios. A large number of the experimental results demonstrate that our algorithm is feasible and efficient for on-road autonomous driving. Furthermore, the comparative test and statistical analyses illustrate that its excellent performance is superior to other previous algorithms.

An improved RRT-based motion planner for autonomous vehicle in cluttered environments.

In this paper, we present an improved RRT-based motion planner for autonomous vehicles to effectively navigate in cluttered environments with narrow passages. The planner first presents X-test that can identify passable narrow passages, and then perform an efficient obstacles-based extension operation within passable narrow passages. In order to generate a smooth trajectory for the vehicle to execute, a post-process algorithm with trajectory optimization is proposed. For the purpose of demonstrate benefits of our method, the proposed motion planner is implemented and tested on a real autonomous vehicle in cluttered scenarios with narrow passages. Experimental results show that our planner achieves up to 13.8 times and 7.6 times performance improvements over a basic RRT planner and a Bi-RRT planner respectively. Moreover, the resulting path of our planner is more smooth and reasonable.

A Multiple Attribute-based Decision Making model for autonomous vehicle in urban environment

In this paper, a maneuver decision making method for autonomous vehicle in complex urban environment is studied. We decompose the decision making problem into three steps. The first step is for selecting the logical maneuvers, in the second step we remove the maneuvers which break the traffic rules. In the third step, Multiple Attribute Decision Making (MADM) methods such as Analytic Hierarchy Process (AHP) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) are used in the process of selecting the optimum driving maneuver in the scenario considering safety and efficiency. AHP is used for obtaining the weights of attributes, TOPSIS is responsible for calculating the ratings and ranking the alternatives. Road test indicates that the proposed method helps the autonomous vehicle to make reasonable decisions in complex environment. In general, the experiment results show that this method is efficient and reliable.

An intent inference based dynamic obstacle avoidance method for intelligent vehicle in structured environment

In this paper, a novel intent inference based dynamic obstacle avoidance method is described that is effective for both normal moving obstacles and maneuvering obstacles, therefore taking advantages of information from different sensors and electronic map. Unlike many of the methods discussed in literature our approach is based on inferring the intent of moving obstacles. It avoids the prediction of accurate maneuvering trajectories which usually fails as the motion is nonlinear and difficult to be modeled. With this method, the performance of dynamic obstacle avoidance is improved which enables the intelligent vehicle to navigate autonomously among moving obstacles. A linear model is used when no maneuver intent is detected and, otherwise, the intent of changing lane is inferred by combing its dynamic features and the road structure when the obstacle is maneuvering. Then a potential collision point will be worked out and avoided to ensure the intelligent vehicle collision free. Experiments show that our method bears significant performance.

UGV Robust Path Following Control under Double Loop Structure with μ Synthesis

For the problem that the path following performance is degraded due to model uncertainty of the unmanned ground vehicle(UGV) during lateral maneuvering,a double loop control structure is designed,in which the path following control is the external loop and the yaw stability control is the inner loop respectively.A robust yaw stability control based on μ synthesis is proposed.Simulation results show that the UGV based on this method has better performance than PID(proportional-integral-derivative) and H∞ controller when model parameters are changed.In comparison experiments,the root mean square error of this method is 1/3 less than PID.The result shows that this method also has robust stability and robust performance with respect to uncertain vehicle parameters.