Trajectory Planning For Autonomous Exploration Of Multirotor MAV In The Unknown Area

The use of multi-rotor micro air vehicle(MAV)to explore unknown areas has broad application prospects in the military and civilian fields,among which exploration trajectory planning is an important area of research.Exploration trajectory planning usually consists of two parts – exploration task planning and corresponding trajectory planning.Generally,exploration task planning provides MAVs with targets,and trajectory planning creates an obstacle-free flight.The goal of both planning steps is to improve the exploration efficiency of the MAV.This paper studies the exploration trajectory planning of a single MAV and multiple multi-rotor MAVs,respectively.For a single MAV,the difficulty of trajectory planning is not only in avoiding obstacles,but also in meeting the dynamic constraints of the MAV,adapting to a large exploration area,and being able to operate online.For multiple MAVs,the difficulty of the exploration task planning is in maximizing its efficiency while avoiding collisions between any two MAVs.The difficulty of trajectory planning lies in avoiding collision with other MAVs,and both planning steps need to be implemented in a decentralized manner.This paper studies the difficulties in planning the exploration trajectory of a single MAV and multiple multi-rotor MAVs.1.To improve a single MAV’s ability to deal with large-scale maps and complex environments,this paper proposes a method of hierarchical planning of trajectory based on target points.This method uses target points to represent the exploration task and employs a concept of layers to divide all target points into global,intermediate,and local.The global target points are extracted from the exploration boundary points of the map.The goal is to improve the exploration efficiency of the MAV and reduce invalid flights.The MAV conducts coarse-grained trajectory planning on the map according to the global target points to obtain an obstacle-free flight trajectory,which is formed by connecting intermediate target points in a series.Based on the detection range of the front sensor,a local target point can be reasonably constructed from the reachable range of the local trajectory plan to further improve the exploration efficiency.Finally,a non-linear programming method is used to obtain a local trajectory that satisfies the dynamics of the MAV.The effectiveness of this method is verified with simulation experiments.2.Cooperating MAVs also use target points to describe their exploration missions.The goal of planning an exploration mission is to use the shortest path in exploring all target points.The method proposed in this paper transforms this problem into a minimum-maximum open-circuit multi-station multi-traveling salesman problem.First,a hybrid discrete particle swarm optimization algorithm is designed to solve this combinatorial optimization problem,and a variable neighborhood descent algorithm is used to perform a global and local search.Second,to create cooperative problem-solving,the particle swarms scattered in the various drones are connected through local communication using the consensus protocol that redesigns the update formula of the particle’s position and velocity.Numerical experiments verify the effectiveness of this method.3.To avoid multi-MAV collisions,it is usually necessary to introduce a large number of constraints in the cooperative trajectory planning.A sampling-based trajectory planning method is adopted in this paper to improve the success of local trajectory planning.However,the sampling-based method is less time efficient than the methods based on mathematical programming.To improve its time efficiency,this research proposes a trajectory re-planning method based on the fixed node optimal rapid expansion random tree(RRT*FN),RRT*FN-Replan.RRT*FN can maintain a fast-expanding random tree in the solution so that the number of nodes remains unchanged and the same tree can be reused in multiple solutions.On this basis,a strategy of rationally reusing old trees is proposed,so that reliance on old information and exploration of new information can be balanced in the re-planning.The effectiveness of the method is verified with simulation experiments.4.If the MAV plans its trajectory in a three-dimensional space with a sampling-based method,the result is a series of track points that do not indicate the speed and acceleration of the MAV on them.Such a trajectory may not meet the dynamic constraints of the MAV.The method proposed in this paper is based on an efficient trajectory planning method of the motion primitive generator,so that the RRT*method can generate a trajectory that conforms to the dynamics of the MAV and can stop the drone at the target points as needed.In this paper,the method of numerical fitting is used.Based on the constraints of the MAV’s maximum speed and acceleration,a feasible range of movement is calculated,where RRT* expands new nodes for the tree.This ensures that the generated trajectory meets the dynamic constraints of the MAV.The effectiveness of this method is verified with simulation experiments.5.Collision avoidance between MAVs needs to be considered in cooperative trajectory planning.In this paper,a decentralized algorithm for solving combinatorial optimization problems is applied to planning a multi-MAV collision-free trajectory.Based on this decentralized algorithm,the MAV negotiates with its neighbors and avoids generating trajectories that interfere with them.The proposed RRT*-based kinodynamic trajectory planning algorithm is used to generate an obstacle-avoidance trajectory for a single MAV.The MAV and its neighbors transfer trajectory schemes to each other,extracting track points from the schemes,constructing virtual obstacles at these positions,and generating trajectories to avoid a collision.To improve collision avoidance,this paper proposes a method based on the maximum clique algorithm that can fix the trajectory of some drones in the system and modify the trajectory of others.The effectiveness of the algorithm is tested in a simulation environment,and all experiments demonstrate a successful avoidance of the collisions.