基于PX4的地面无人车避障系统及路径规划研究

地面无人车避障及路径规划是指,无人车在自动巡航过程中,遇到障碍物能够自动避开,并对当前行进路线进行重新规划,重新进入预设路线行驶。本文使用HPI公司的Savage Flux 2350差速驱动越野车作为实验平台,选用开源硬件Pixhawk飞控作为运动控制器,对PX4 Rover软件系统进行二次开发。使用数据采集模块获得超声波数据,并通过串口方式与飞控板进行通信,完成数据交换,从而实现对地面无人车周围环境的实时检测,完成巡航模式下的避障功能。

基于视觉显著性的无人车图像检测及分割方法

障碍物检测与分割是地面无人车辆环境感知领域中一项重要的任务。针对传统障碍物检测与分割算法的计算量大、分割精度较差等问题,提出了一种基于显著性分析的障碍物检测、分割优化算法。首先,利用基于频率调谐的方法生成场景图像的显著图;然后,通过单目摄像机与激光雷达的联合标定将雷达反射点映射到显著图上;最后,结合单目摄像机和激光雷达两种传感器信息,通过改进的图像区域分割算法,实现障碍物的检测与分割。为了验证所提出算法有效性,采集多幅包含障碍物的典型越野场景图像,对该算法进行实验与仿真验证,结果证明了该算法的有效性。

基于多模态感知与融合的无人车韧性导航系统

针对野外复杂环境下的无人车自主导航需要,建立了一种基于多源融合定位、语义建图与运动规划的智能导航系统。首先,针对IMU、轮式里程计、视觉SLAM与激光雷达SLAM等测量子系统,设计了误差状态扩展卡尔曼滤波器进行融合定位。其次,基于改进的CNN语义分割网络生成环境的语义图像,与3D激光雷达点云融合,并使用最大概率更新算法构建语义3D地图。接着,在语义和几何信息投影获得可通行性代价的基础上,提出了一种语义动态窗口的局部路径规划方法。最后,将以上感知、定位与规划方法整合成完整的智能导航系统,在城市与野外典型场景的测试中,相对定位误差小于0.4%D,具备一定的韧性导航定位和智能感知规划能力。

Development and Missions of Unmanned Surface Vehicle

The navy and other Department of Defense organizations are increasingly interested in the use of unmanned surface vehicles (USVs) for a variety of missions and applications. The term USV refers to any vehicle that operates on the surface of the water without a crew. USVs have the potential, and in some cases the demonstrated ability, to reduce risk to manned forces, provide the necessary force multiplication to accomplish military missions, perform tasks which manned vehicles cannot, and do so in a way that is affordable for the navy. A survey of USV activities worldwide as well as the general technical challenges of USVs was presented below. A general description of USVs was provided along with their typical applications. The technical challenges of developing a USV include its intelligence level, control, high stability, and developmental cost reduction. Through the joint efforts of researchers around the world, it is believed that the development of USVs will enter a new phase in the near future, as USVs could soon be applied widely both in military and civilian service.

Trajectory tracking control for underactuated unmanned surface vehicles with dynamic uncertainties

The trajectory tracking control problem for underactuated unmanned surface vehicles(USV) was addressed, and the control system took account of the uncertain influences induced by model perturbation, external disturbance, etc. By introducing the reference, trajectory was generated by a virtual USV, and the error equation of trajectory tracking for USV was obtained, which transformed the tracking problem of underactuated USV into the stabilization problem of the trajectory tracking error equation. A backstepping adaptive sliding mode controller was proposed based on backstepping technology and method of dynamic slide model control. By means of theoretical analysis, it is proved that the proposed controller ensures that the solutions of closed loop system have the ultimate boundedness property. Simulation results are presented to illustrate the effectiveness of the proposed controller.

Adaptive Resource Allocation Algorithm for Internet of Things with Bandwidth Constraint

In order to improve the transmission accuracy and efficiency of sensing and actuating signals in Internet of Things （loT） and ensure the system stability, an adaptive resource allocation algorithm is proposed, which dynami- cally assigns the network bandwidth and priority among components according to their signals＇ frequency domain characteristics. A remote sensed and controlled unmanned ground vehicle （UGV） path tracking test-bed was devel- oped and multiple UGV＇s tracking error signals were measured in the simulation for performance evaluation. Results show that with the same network bandwidth constraints, the proposed algorithm can reduce,, the accumulated and maximum errors of UGV path tracking by over 60% compared with the conventional static algorithm.

Serret-Frenet frame based on path following control for underactuated unmanned surface vehicles with dynamic uncertainties

The path following problem for an underactuated unmanned surface vehicle(USV) in the Serret-Frenet frame is addressed. The control system takes account of the uncertain influence induced by model perturbation, external disturbance, etc. By introducing the Serret-Frenet frame and global coordinate transformation, the control problem of underactuated system(a nonlinear system with single-input and ternate-output) is transformed into the control problem of actuated system(a single-input and single-output nonlinear system), which simplifies the controller design. A backstepping adaptive sliding mode controller(BADSMC)is proposed based on backstepping design technique, adaptive method and theory of dynamic slide model control(DSMC). Then, it is proven that the state of closed loop system is globally stabilized to the desired configuration with the proposed controller. Simulation results are presented to illustrate the effectiveness of the proposed controller.

Modeling and control for hydraulic transmission of unmanned ground vehicle

Variable pump driving variable motor(VPDVM) is the future development trend of the hydraulic transmission of an unmanned ground vehicle(UGV).VPDVM is a dual-input single-output nonlinear system with coupling,which is difficult to control.High pressure automatic variables bang-bang(HABB) was proposed to achieve the desired motor speed.First,the VPDVM nonlinear mathematic model was introduced,then linearized by feedback linearization theory,and the zero-dynamic stability was proved.The HABB control algorithm was proposed for VPDVM,in which the variable motor was controlled by high pressure automatic variables(HA) and the variable pump was controlled by bang-bang.Finally,simulation of VPDVM controlled by HABB was developed.Simulation results demonstrate the HABB can implement the desired motor speed rapidly and has strong robustness against the variations of desired motor speed,load and pump speed.

Trophallaxis network control approach to formation flight of multiple unmanned aerial vehicles

A novel network control method based on trophaUaxis mechanism is applied to the formation flight problem for multiple un- manned aerial vehicles （UAVs）. Firstly, the multiple UAVs formation flight system based on trophallaxis network control is given. Then, the model of leader-follower formation flight with a virtual leader based on trophallaxis network control is pre- sented, and the influence of time delays on the network performance is analyzed. A particle swarm optimization （PSO）-based formation controller is proposed for solving the leader-follower formation flight system. The proposed method is applied to five UAVs for achieving a ＇V＇ formation, and a series of experimental results show its feasibility and validity. The proposed control algorithm is also a promising control strategy for formation flight of multiple unmanned underwater vehicles （UUVs）, unmanned ground vehicles （UGVs）, missiles and satellites.

An improved artificial physics approach to multiple UAVs/UGVs heterogeneous coordination

This paper proposed an improved artificial physics(AP)method to solve the autonomous navigation problem for multiple unmanned aerial vehicles(UAVs)/unmanned ground vehicles(UGVs)heterogeneous coordination in the three-dimensional space.The basic AP method has a shortcoming of easily plunging into a local optimal solution,which can result in navigation fails.To avoid the local optimum,we improved the AP method with a random scheme.In the improved AP method,random forces are used to make heterogeneous multi-UAVs/UGVs escape from local optimum and achieve global optimum.Experimental results showed that the improved AP method can achieve smoother trajectories and smaller time consumption than the basic AP method and basic potential field method(PFM).