✨ 长期致力于路径规划、RRT~*算法、人工势场法、自动巡检研究工作擅长数据搜集与处理、建模仿真、程序编写、仿真设计。✅ 专业定制毕设、代码✅如需沟通交流点击《获取方式》1提出基于安全边界与朝向合力场随机游走的改进RRT*算法传统RRT*在处理不规则障碍物时路径易穿透障碍物边界。改进方法在每个障碍物周围增加安全边界边界宽度为机器人半径零点三米的两倍。在采样阶段引入朝向合力场方向的随机游走合力场由目标引力与障碍物斥力合成引力系数零点六斥力系数零点四。当随机游走步数超过十步仍未找到可行节点时回退到均匀采样。采用Lin-Kernighan算法对生成的路径进行局部优化选取最优路径。在Matlab仿真中设置五个随机凸多边形障碍物传统RRT*平均路径长度十二点四米改进算法十点一米规划时间从三点二秒降至一点五秒。路径平滑度通过曲率积分评估改进算法降低百分之三十。2设计改进人工势场法解决目标不可达与局部最优问题斥力势场函数引入调节因子dist(q,q_goal)^nn取值为二使得目标点处斥力为零。引力势场函数增加范围限定dg*设dg*为两米当机器人与目标距离大于两米时引力保持恒定最大值避免引力过大导致碰撞。同时增加速度斥力势场速度斥力系数设为五使机器人在面对动态人流时减速。仿真对比中传统人工势场法在狭窄通道中陷入局部最优概率百分之三十五改进后降至百分之五。在含三个凹形障碍物的地图中改进算法成功率达到百分之九十八路径长度平均十五点七米。3融合改进RRT*与改进人工势场法的全局-局部混合路径规划系统系统工作流程首先使用改进RRT*生成全局粗糙路径路径点间距零点五米。机器人行驶时实时检测前方零点八米范围内是否出现未建模动态障碍物。若触发避障条件切换至改进人工势场法进行局部避障待绕过障碍物后重新接回全局路径。使用Bezier曲线对全局路径进行二次平滑处理控制点取相邻三个路径点曲线次数为三。在ROS Gazebo中搭建会场仿真环境包含展台、立柱与模拟行人。融合算法平均每十次试验中成功完成九点七次路径跟随误差平均零点零八米。搭建真实巡检机器人试验平台配置激光雷达RPLIDAR A1与STM32控制器。在二十米乘二十米的会场区域测试融合算法路径规划用时零点三秒比单一RRT*快百分之六十转向角变化率每秒小于十五度满足舒适性要求。import numpy as np import random class ImprovedRRTstar: def __init__(self, start, goal, obstacles, safe_margin0.3): self.start np.array(start) self.goal np.array(goal) self.obstacles obstacles # list of (center, radius) self.safe_margin safe_margin self.tree [self.start] self.cost {tuple(self.start): 0} def potential_field_direction(self, pos): F_att 0.6 * (self.goal - pos) F_rep np.zeros(2) for center, rad in self.obstacles: d np.linalg.norm(pos - center) if d rad self.safe_margin: F_rep 0.4 * (1/(d1e-6)) * (pos - center) / d return F_att F_rep def sample_with_bias(self): if random.random() 0.7: # 偏置采样 direction self.potential_field_direction(self.tree[-1]) if np.linalg.norm(direction) 1e-3: direction direction / np.linalg.norm(direction) return self.tree[-1] direction * random.uniform(0.2, 1.0) return np.random.rand(2) * 10 def extend(self, max_iter500): for _ in range(max_iter): q_rand self.sample_with_bias() q_near min(self.tree, keylambda x: np.linalg.norm(x - q_rand)) step 0.3 q_new q_near step * (q_rand - q_near) / max(np.linalg.norm(q_rand - q_near), 1e-6) if not self.collision_free(q_near, q_new): continue self.tree.append(q_new) self.cost[tuple(q_new)] self.cost[tuple(q_near)] np.linalg.norm(q_new - q_near) if np.linalg.norm(q_new - self.goal) 0.5: return True return False def collision_free(self, p1, p2): for center, rad in self.obstacles: v p2 - p1 w center - p1 t np.dot(w, v) / np.dot(v, v) t max(0, min(1, t)) closest p1 t * v if np.linalg.norm(closest - center) rad self.safe_margin: return False return True class ImprovedAPF: def __init__(self, start, goal, obstacles, dg_star2.0, n2): self.pos np.array(start) self.goal np.array(goal) self.obstacles obstacles self.dg_star dg_star self.n n def attractive(self): dist np.linalg.norm(self.pos - self.goal) if dist self.dg_star: return self.dg_star * (self.goal - self.pos) / dist else: return (self.goal - self.pos) def repulsive(self): force np.zeros(2) for center, rad in self.obstacles: d np.linalg.norm(self.pos - center) if d rad: # 调节因子 (dist to goal)^n dist_g np.linalg.norm(self.pos - self.goal) force 0.5 * (1/d - 1/rad) * (self.pos - center) / (d**3) * (dist_g**self.n) return force def step(self, step_size0.1): F self.attractive() self.repulsive() self.pos step_size * F / (np.linalg.norm(F)1e-6) return self.pos if __name__ __main__: obs [((3,3),0.5), ((5,6),0.4), ((7,2),0.6)] rrt ImprovedRRTstar(start(0,0), goal(9,9), obstaclesobs) success rrt.extend(max_iter300) print(fRRT* success: {success}, tree size: {len(rrt.tree)}) apf ImprovedAPF(start(0,0), goal(9,9), obstaclesobs) path_apf [apf.pos.copy()] for _ in range(100): new_pos apf.step() path_apf.append(new_pos.copy()) if np.linalg.norm(new_pos - apf.goal) 0.2: break print(fAPF路径长度: {len(path_apf)})