Research On Intelligent Car-like Robots Collision Avoidance

In recent years, because of the rapidly development of artificial intelligence, intelligentrobot is gaining more and more attention. As a member of intelligent robot, car-like robot isexpected to solve transportation issues such as traffic jam and energy consumption byself-driving. Car-like robot has huge potential in Intelligent Transportation System (ITS).Vehicular ad-hoc network (VANET) is the core of intelligent transportation system (ITS),VANET is a self-organized network established between moving vehicles and roadsideinfrastructures. VANET can provide communication between vehicles (V2V) and thecommunication between vehicle and roadside node (V2I). Through Cooperative AwarenessMessage (CAM) in V2V communications, vehicles can share their status information to eachother (speed, position, yaw angle, etc.). Compared to the traditional sensors, method utilizinginter-vehicle communication provides stronger perception ability. In this thesis, amulti-vehicle collision avoid algorithm (MVCA) is designed, also the influences of broadcastdelay and packet loss to trajectory planning is uncovered.The main work of this thesis is as follows:(i) A global navigation algorithm for intelligent vehicle is designed based on A*, whichindicate car-like robot a global path to destination.(ii) A real-time distributed multiple vehicle collision avoid algorithm (MVCA) isproposed on the basis of Reciprocal n-body Collision Avoidance method. The advantage ofthis algorithm is to allow each robot decides their destination and control input independently,without the negotiation with each other or the coordinator. Compared to some centralizedtrajectory planning algorithm, decentralized approach reduces the compute cost and greatlyimprove the robustness of the system. The destination of each robot can be regarded asprivacy.(iii) The simulation is conducted on MATLAB and C++platform, the MVCA algorithmis validated on typical intersection scenarios. The destination of each car-like robot is the onlyrequirement, simulation results show that the algorithm we proposed is safe and reliable.(iv) Although VANET has its huge potential, related works on the influences of latencyand packet loss to trajectory planning is still rare. The MVCA algorithm is simulated underthe vehicular ad-hoc network (VANET), the influence introduced by communication delay and packet loss is uncovered.