Research On Robot Behavior Strategy In Emergency Situations

When the mobile robot is working in the field,it is hard to avoid all kinds of faults.The robot vision system is the main way to measure the external environment.When it is contaminated or even damaged,the robot could be paralyzed because it cannot correctly perceive the external environment.To solve this,this paper mainly studies the contingency planning design based on case-based reasoning(CBR)under the emergency situation,so as to enable the robot have the degraded continuous sense ability and the motion control adapt to rugged road.And a control method based on neural network controlling the tractive force of robot to go through the rugged road was proposed.On a ground-adaptive wheeled mobile robot platform with self-cleaning capabilities,a case-based reasoning(CBR)contingency plan for robot failure was designed,and a contingency planning design was established.The kinematics model of the robot platform was established.And aiming at obstacle avoidance of robots,an obstacle avoidance algorithm based on sub-object point was presented.To achieve a degraded environment awareness for robots,a behavioral strategy which applied the two-dimensional(2D)lidars,used to measure grounding angles,to implement the sensor modelling of three-dimensional(3D)terrain was proposed in this paper.The identification of obstacle and the establishment of the ground model were accomplished based on radar data.The purposes of 3D terrain perception are listed as follows: Identification of the ground obstacle;Visualized modeling of the ground surface;Digital elevation map(DEM)establishment of the frontal terrain.In order to improve the ability of the robot to pass rugged road surface,a traction control strategy based on neural network is proposed.The terrain DEM was constructed by lidars.The simulation estimation algorithm of the posture and force distribution is used to obtain the robot posture and force distribution simulation data under the scanning terrain based on multi-objective optimization algorithm.The simulation data is taken as the learning samples for neural network.A network model of robot traction control was built.And a speed detection algorithm based on 2D lidars is presented.Taking the robot platform under study as the experimental object,the terrain scanningperceptual experiment is carried out to construct the ground model.The control experiment of robotic traction network was accomplished on symmetrical surfaces,asymmetric surfaces and rugged roads.The motor data of drive wheels about the current,speed and track angle fed back by the motor drier was analyzed.And the feasibility of the control method of traction force network was testified.