전방 선회를 고려한 무인자율주행차량의 지역경로계획과 경로추종

Abstract

미래형 자동차를 위한 기술 개발은 저탄소, 고연비 조건을 만족시키는 친환경 기술과 더불어 운전자의 안전성과 편의성을 극대화 할 수 있는 지능형 기술에 집중되고 있다. 그 중 무인자율주행차량(Autonomous Vehicles)은 군사적 목적으로 연구를 진행하여 미 국방 고등연구 기획청에서 주최한 2004, 2005 DARPA(The Defense Advanced Research Projects Agency) Grand Challenge를 거치면서 두각을 나타내어 2007 DARPA Urban Challenge를 통하여 60마일(96km)을 완주하는 성과를 기록하며 급속한 발전을 하였다. DARPA Urban Challenge에 참여한 Stanford University에서 소개한 Stanley Steering Controller는 간단한 수식과 우수한 경로추정 성능을 가진다. 하지만 Stanley Steering Controller의 거리오차와 각도오차는 조향 이득값(Gain)에 의해 크게 영향을 받는다. 여기서 거리오차는 기준경로에서 차량이 떨어진 정도이며, 각도오차는 기준경로의 진행방향과 자동차의 방향각간의 편차를 나타낸다. 또한, 이 Controller는 경로에 수직인 거리오차를 줄이는 방법으로, 경로계획을 정확하게 할 경우 최대의 성능을 가진다는 단점이 있다. 본 논문에서는 GPS 수신기를 이용한 경로의 경유점(Waypoint) 생성과 목표 횡가속도(Lateral Acceleration)를 이용한 새로운 목표점 생성방법과 경로추종방법을 제안한다. 새로운 경로추종방법을 통해 주행시간과 주행거리를 단축시키는 것은 물론, 측방향 거리오차를 줄이며, 교차로 주행시 Stanley 경로추종에서 발생하는 Curve Overshoot을 줄여 부드러운 선회가 가능하게 한다.|Path planning methods including A*, D*, Dijkstra, and RRTs (Rapidly-exploring Random Trees) are introduced through DARPA Urban and Grand challenges. Moreover, the path following methods including Stanley, Pure Pursuit, EMPC (Error Model-based Predictive Control) are also presented. In this paper, new path following method is proposed. This method has simple expression as well as excellent performance, compared with the other path following methods. When the detailed path planning deos not provided, the vehicle has a large curve overshoot during the drive. To deal with this problem, the path following method is proposed to reduce the curve overshoot and dependence of detailed path planning. Proposed path following method uses the lateral acceleration principle, which is proposed for UAV (Unmanned Aerial Vehicle). This method sets a virtual line from the current vehicle position to the target point on the path. Then, the radius of curvature is generated using the length of virtual line. Using proposed path following method, the numerical simulation is performed. In addition, the simulation results are compared with those of Stanley method, and then it is shown that the proposed method has less curve overshoot than other method to track the designed path. Therefore, through the proposed method, the autonomous vehicle can drive smooth tracking in a curved path.; Path planning methods including A*, D*, Dijkstra, and RRTs (Rapidly-exploring Random Trees) are introduced through DARPA Urban and Grand challenges. Moreover, the path following methods including Stanley, Pure Pursuit, EMPC (Error Model-based Predictive Control) are also presented. In this paper, new path following method is proposed. This method has simple expression as well as excellent performance, compared with the other path following methods. When the detailed path planning deos not provided, the vehicle has a large curve overshoot during the drive. To deal with this problem, the path following method is proposed to reduce the curve overshoot and dependence of detailed path planning. Proposed path following method uses the lateral acceleration principle, which is proposed for UAV (Unmanned Aerial Vehicle). This method sets a virtual line from the current vehicle position to the target point on the path. Then, the radius of curvature is generated using the length of virtual line. Using proposed path following method, the numerical simulation is performed. In addition, the simulation results are compared with those of Stanley method, and then it is shown that the proposed method has less curve overshoot than other method to track the designed path. Therefore, through the proposed method, the autonomous vehicle can drive smooth tracking in a curved path.